![Chapter 1
A Community-based Approach for Service-based
Application Composition in an Ecosystem
Elie Abi-Lahoud, Marinette Savonnet, Marie-Noelle Terrasse, Marco Viviani,
Kokou Yétongnon
Université de Bourgogne – Sciences et Techniques, Laboratoire LE2I – Mirande,
Aile de l’Ingénieur, 9, av. Savary, 21078 Dijon cedex, France
The design of composite applications by combining existing services with known seman-
tics is an ongoing topic in current research. Several studies are aimed at providing service
description models and standards, service discovery and matching etc. However, service
composition in distributed dynamic environments such as P2P ecosystems has received
little attention from research communities. In this paper we present a design framework
for composing services, taking in particular into account different ways of building peer-
communities based on network or services characteristics.
1.1 Introduction
Service oriented computing provides software designer with new concepts and emerging
principles for developing loosely-coupled, cross-enterprise business applications. Tradi-
tionally, software development approaches rely on CASE tools [1] and modeling concepts
to describe and implement software components that can be integrated into applications.
Recently, we are witnessing a shift from this static view of software development and de-
ployment towards a dynamic, adaptable service based view of software design in which
applications could be realized in a flexible manner to respond to changing needs of users.
In this emerging design view, services provide high level functional components that can
be shared in open distributed environments. The goal is to design composite applications
by combining existing service components with known semantics, spanning organizations
and computing platforms.
Many research efforts have been aimed at service oriented computing, ranging from tech-
A. Gabillon et al., Web-Based Information Technologies and Distributed Systems, 1
Atlantis Ambient and Pervasive Intelligence 2, DOI 10.2991/978-94-91216-32-9_1,
© 2010 Atlantis Press/World Scientific](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-21-320.jpg)
![2 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon
nical services to telecommunication services, business process modeling and popular web
services. In the information system realm, this research effort has focused to a significant
extent on (i) services definition: linguistic constructs and models to define and represent
services’ behaviors and properties, (ii) services discovery: architectures or protocol suites
to allow service sharing and functional matching and (iii) services composition: orchestra-
tion of service components into more complex processes [2, 3, 4, 5].
Open computing environments created the needs for virtual cooperating systems to allow
resource sharing. Digital enterprise ecosystems emerged as a concept for capturing the in-
teractions of business networks. Ecosystems can comprise autonomous organizations and
related services, sharing agreements on overall domain specific components and rules gov-
erning interactions and inter-relationships among the participants. Enterprise ecosystems
provide some formalization of common models, shared knowledge and global resources to
enable loosely coupled interoperability among enterprises. In essence, they can be used,
as opposed to open environments, to provide controlled business and enterprise environ-
ments delimiting the collaboration scopes to a set of actors respecting business related
rules. Ecosystems form a suitable environment for application composition. They provide
an environment with identified semantics and business properties wherein peers providing
services interact based on a global but not too restrictive agreement. This helps in distin-
guishing functional needs, relations between them and other business-relevant properties.
1.1.1 Objectives and Contributions
In this chapter we address the service-based application composition issue in a peer-to-
peer ecosystem. In such an ecosystem, our approach consists of first defining a high level
interaction between actors, then refining it to an application defined as a graph of abstract
services. The application is realized by substituting abstract services for matching services
provided by peers belonging to the ecosystem. We show how the concrete application real-
ization based on service composition can take advantage of the ecosystem’s network reor-
ganization into peer communities, in terms of communities’ definition and communication
protocol by building on top of an unstructured system a hybrid overlay network.
The remainder of the chapter is organized as follows. Section 1.2 exposes literature back-
ground, namely service orientation and P2P systems and communities. Section 1.3 exposes
ecosystems, peer-communities and services under a multi-layered comprehensive frame-
work for service-based application composition. Section 1.4 focuses on the fourth layer
of the framework, describing its organization in a super-peer based overlay network. Sec-](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-22-320.jpg)
![A Community-based Approach for Service-based Application Composition in an Ecosystem 3
tion 1.5 presents the European Electricity Exchange Market as an ecosystem example. It
compares two views of the studied ecosystem, focusing on the process of application real-
ization. Section 1.6 concludes the chapter and presents future work.
1.2 Background
In this Section we discuss two recent developments that are changing the way IT appli-
cations are designed, deployed and exchanged: (i) service oriented computing, providing
a new paradigm for creating applications on demand and (ii) peer-to-peer systems, often
used for sharing resources. We first describe current work in service oriented computing,
then we briefly define P2P systems and review P2P communities-related literature.
1.2.1 Service Orientation
Previous work in service oriented systems has focused to a significant extent on 1) con-
structs and models to define and represent the behaviors and properties of services, and
2) the architectures or protocol suites to allow service sharing and matching and on ser-
vices’ composition into more complex systems.
A service can be viewed as a self-contained, modular basic software unit that is described,
published and invoked over a network to create new software components or products. It
encapsulates functions and modules of an application domain (e.g., business process com-
ponents, supply chain units). It provides an interface to allow external invocation. Among
service description models proposed in the literature, the Web Service Description Lan-
guage (WSDL) [2] has become a de-facto industry standard. It is an XML-based model
that allows a syntactical representation of the methods and parameters needed to inter-
act with a service. Other models extend the syntactic representation of services by adding
semantics to resolve definition discrepancies and heterogeneities that can hinder service
matching and composition. For example, the METEOR-S project [3] extend WSDL with
semantic annotations while the OWL-S [4] and WSMO [5] (the Web Service Modeling
Ontology) approaches are based on an ontology of web services. The ontology provides a
precise description of service components and their inter-relationships. Several standards
and architectures are proposed to enable the integration and sharing of heterogeneous ser-
vice. For example, Service Oriented Architecture (SOA) is a “paradigm for organizing and
utilizing distributed capabilities that may be under the control of different ownership do-
mains” [6, 7].](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-23-320.jpg)
![4 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon
Service discovery is defined by Keller et al. [8] as the automatic localization of services cor-
responding to user’s need. Booth et al. [9] describe the discovery process as the localization
of a machine readable description corresponding to given functional needs. Toma et al. [10]
define service discovery as a process taking as input a user query and returning as output
a list of available resources corresponding to the user’s need expressed in the input query.
Two major aspects are tackled by service discovery, namely service localization and ser-
vice matching. Service localization relies on either centralized or distributed architectural
models. The UDDI (Universal Description Discovery and Integration [11, 12, 13]) became
a widely know standard for centralized service localization. It consists of a set of UDDI
nodes collaborating to create a global structure. Srinivasan et al. [14] extended the UDDI
model to support OWL-S semantic description allowing more efficient comparison between
the user’s need and the available services. Distributed localization models consisted first in
setting up a distributed federation of UDDIs [15, 16]. Verma et al. [17], Paolucci et al.
[18] and Schmidt et al. [19] discussed other complex models. Service matching is widely
addressed in the literature. Ernst et al. [20] and Dong et al. [21] studied syntactic similarity
based on trace data and clustering respectively. Paolucci et al. [22], Benatallah et al. [23]
and the WSMO workgroup [8] tackled the matching based on semantic similarity. Taher et
al. [24] and Bordeaux et al. [25] studied other approaches based on abstract services and
labeled transition systems respectively.
Service composition designates the interaction taking place between two or more services
in order to accomplish a given goal. The composition process tackles several aspects, such
as the interaction description and organization, the message exchange management, the
transaction like behavior, the interaction context, the level of automation, the failure recov-
ery, etc. The Web Services Business Process Execution Language (WS-BPEL [26]) is the
current standard for describing services’ compositions. It allows to model compositions as
interaction workflows. An alternative for BPEL is the Web Service Choreography Interface
(WSCI [27]). Both BPEL and WSCI allow static service composition wherein services are
bound at design time. Thakkar et al. [28], Casati et al. [29] and Sun et al. [30] present
dynamic composition environments based on composition engines capable of binding se-
lected services at runtime. The automation level of the composition process is also widely
studied in the literature. An exhaustive survey on service composition is out of the scope of
this work. Useful information is available in [31, 32].](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-24-320.jpg)
![A Community-based Approach for Service-based Application Composition in an Ecosystem 5
1.2.2 P2P Systems
Peer-to-peer (P2P) systems are distributed systems composed of distinct computing ele-
ments, called peers, with similar resources and capabilities. Peers interact together to share
services and resources. P2P systems can be classified into unstructured and structured sys-
tems. In unstructured P2P systems, peers are organized in random graphs with no control
over their contents. Each peer controls its contents and the access and sharing of its re-
sources. Unstructured P2P systems can be further classified into (i) centralized systems
when a central directory is used to store global state information (indexes, data locations,
etc.), (ii) decentralized systems when no global state information (network state informa-
tion, context data) is maintained by the peers, and (iii) hybrid systems which combine the
characteristics of centralized and decentralized by using super-nodes (or super-peers) [33]
to control simple peers with less resources and capabilities. Structured P2P systems keep a
tight control over network topology and peer contents by placing data not randomly in peers
but at specific locations defined by the overlay network strategy (an indexing strategy).
P2P systems can also be structured by using clustering techniques to group peers based
on common properties or interests. Clusters can be viewed as communities belonging to
overlays defined on top of unstructured P2P systems. According to Khambatti et al. [34],
a community is a set of active peer members, involved in sharing, communicating and pro-
moting common interests. Significant research is currently targeted at creating community-
oriented overlay networks in order to avoid query messages flooding and to save resources
in handling irrelevant queries over the P2P network. DHT-based techniques [35, 36] guar-
antee location of content within a bounded number of hops by tightly controlling the data
placement. Other techniques based on clustering strategies have been proposed to reduce
query traffic, grouping peers sharing similar properties.
According to Oztopra et al. [37], two main strategies are used the literature for clustering
peers. The first strategy takes into account network related characteristics while the second
focuses on peers interests. In the following we review both strategies considering that peers
participating in a services’ ecosystem are mainly interested in providing, sharing and re-
using services.
1.2.2.1 Using Network Characteristics to Build Peer Communities
Several research was conducted on clustering of peers based on network characteristics.
Ratnasamy et al. [38] present a scheme whereby nodes partition themselves into groups
called bins such that nodes that fall within a given bin are relatively close to one another in](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-25-320.jpg)
![6 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon
terms of network latency. Zhang et al. [39] propose a topology aware system constructing
an overlay network by exploiting the locality in the underlying network using the group
concept. Each host in the overlay is running a protocol to communicate with other hosts.
In general, each host maintains information about a set of other hosts to communicate
with. Two hosts are considered as neighbors if they are connected through the overlay.
MetaStream [40] is a content discovery protocol for topology-aware on-demand streaming.
In MetaStream, clients choose streaming sources based on network distance. For this pur-
pose, they self-organize into a dynamic hierarchy of clusters based on the network topology.
Any protocol for constructing a topology-aware hierarchy can be used. Connectivity-based
Distributed node Clustering (CDC) [41] implements node clustering based on node con-
nectivity in P2P networks, while Zheng et al. [42] use an approach based on link delay
of node communications in the P2P network. Oztopra et al. [37] propose to cluster peers
based on time (communication duration) closeness.
Disregarding the specific adopted technique, building communities based on network char-
acteristics generates an overlay where peers in a community provide different services.
In such a scenario, it is highly probable that peers in a community will behave in a co-
operative manner. When one peer is selected, the possibility of selecting another member
of the community is increased. This makes sense considering that it is better for a peer to
search for a service among his neighbors before searching among further members.
1.2.2.2 Using Service Properties to Build Peer Communities
Service properties are classified in two main categories: (i) functional and (ii) non-
functional [5, 43].
Functional properties represent the functionality provided by the service and its semantic
description elements, for example the related input/output parameter list (and conditions if
available). Note that the service as a software unit might provide several functionalities. In
this case, each functional aspect can be studied as a separate entity. Building communities
based on the functionalities provided, allows us to obtain competitive communities, where
each peer holds services accomplishing the same task, although some service attributes
may vary. This way, each peer will compete with others to get selected by a client. The
client choice is based on non-functional properties, which are not directly related to the
functionality provided by the service.
An exhaustive list or classification of those properties is out of the scope of our project, we
note that some of the non-functional properties are QoS related and thus in correlation](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-26-320.jpg)
![A Community-based Approach for Service-based Application Composition in an Ecosystem 7
with the network characteristics such as execution time. Other non-functional properties
do not express QoS but might form substrate criteria to build communities on, for instance
security-level and trust [44, 45].
1.3 A Framework for Sharing Services
In the following we present and model peers interactions in a services’ ecosystem. We de-
fine peer-communities in such ecosystems and formalize interaction rules. We also describe
the multi-layered service-based composition framework under which on-demand applica-
tion composition takes place.
1.3.1 Ecosystem, Peer-communities and Services
In an ecosystem various communities organize business-driven collaboration among groups
of service-providing peers. We first describe such ecosystems, and then we provide more
precise definitions of the relevant terms.
%HORQJVWR %HORQJVWR
3DUWLFLSDWHVLQ
3HHU 1L
0XVWDJUHHZLWKWKH
HFRVVWHPJOREDO
DJUHHPHQW *ORE$ (
3HHURPPXQLW 3/RF$
0XVWDJUHHZLWKWKHHFRVVWHP
JOREDODJUHHPHQW*ORE$ (
0XVWGHILQHDORFDODJUHHPHQW
(FRVVWHP (
0XVWGHILQHDJOREDO
DJUHHPHQW *ORE$ (
0XVWGHILQHDVHWRI
J J
/RF$ 3 DEVWUDFWVHUYLFHV$
6HUYLFHV
3URYLGHV
$EVWUDFW 6HUYLFHV
$FFHVVHGYLD
,PSOHPHQW
'HILQHV
6HUYLFHV $EVWUDFW 6HUYLFHV
S
Figure 1.1 Ecosystem’s Organization
As described in Figure 1.1, an ecosystem is a group of peer-communities in which each
peer-community accepts a consensual specification of a business area and its related busi-
ness rules, referred to as global agreement. The ecosystem defines a set of abstract services
based on the global agreement. A peer respecting the global agreement provides services
as implementations of the abstract ones. Peer-communities are groups of peers having a
consensual agreement on a minimum set of properties. We denote by local agreement this
set of the required properties. Each peer must satisfy its ecosystem’s global agreement and
its communities’ local agreements.](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-27-320.jpg)
![8 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon
We model the ecosystem’s organization by the following definitions and rules:
• Given an ecosystem E, its global agreement specification is the set of business-related
properties denoted by GlobA(E). GlobA(E) = {pp}p=[1,...,q] where q is the number
of relevant business properties in the ecosystem. For example response time, which
is a widely used property. Properties notation language is chosen depending on the
ecosystem domain(s).
• Abstract services are defined by the ecosystem in order to disseminate an application
domain knowledge. An abstract service is an interface defining the abstract operation
needed to fulfill the related functionality. An abstract service describes the operation
via a semantic business-related description, including but not limited to, its inputs,
outputs and an associated set of constraints, typically restrictions. It provides no real
implementation of the operation, just the signature. The abstract service interface al-
lows to define also realization constraints to be respected by the interface implementer.
An abstract service is designated by Ai. We denote by A the set of abstract services de-
fined by the ecosystem E, and by n the number of described abstract services in the
ecosystem such as |A| = n.
• For each abstract service Ai the set of defined properties Pi is defined as: Pi =
{pm
i }m=[1,...,qi],i∈{1,...,n} subject to qi ⩽ q and ∃ f : pm
i → pp. For instance, based on
the business property response time, the abstract service’s interface defines the prop-
erty execution time. We note that some business related properties might not be relevant
for a given abstract service, thus they are not used in its interface.
• Abstract services defined by an ecosystem E must comply with the GlobA(E) specifi-
cation:
∀Ai ∈ A, Pi ⊆ GlobA(E) where i ∈ {1,...,n}
• Services are defined by peers in order to present their business offers in the ecosys-
tem. They are defined with respect to the ecosystem required functionalities, thus a
service is a concrete implementation of an abstract one. A service Sij implementing
an abstract service Ai must redefine the abstract operation and respect all the associ-
ated constraints. Although having similar functional interfaces, two services Sij and
Sij may differ in their non-functional properties.
• For each service Sij the set of service-related properties is derived from the correspond-
ing abstract service properties and is denoted by Pij = {pp
ij}i∈{1,...,n}, p=[1,...,qi], j∈N.
Actually a service redefines and implements the properties of its related abstract ser-
vice. For example, the service redefines the property execution time inherited from its](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-28-320.jpg)
![10 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon
ďƐƚƌĂĐƚ^ĞƌǀŝĐĞƐ
KŶŽƚůŽŐŝĞƐͬ ĞƐĐƌŝƉƚŝŽŶƐ
ƵƐŝŶĞƐƐĐƚŝǀŝƚŝĞƐ
%XVLQHVV
/DHU
ϭ
Ϯ
Ŷ ͙
Đϭ Đϯ
ĐϮ
ĐŶ
KŶŽƚůŽŐŝĞƐͬĞƐĐƌŝƉƚŝŽŶƐ
'W
$SSOLFDWLRQ
/DHU
^ĞƌǀŝĐĞ
ŽŵƉŽƐŝƚŝŽŶ
6HUYLFH
5HSRVLWRU
ŽŽŬŬĞĞƉŝŶŐ ͗
/ϭ /Ϯ /ŵ
͙
,QVWDQFHV
/DHU
ŽŵŵƵŶŝƚLJ
ŝƐƚŝŶŐŽĨ
ĐŽŶĐƌĞƚĞƐĞƌǀŝĐĞƐ
ĨŽƌĞĂĐŚ/ŶƐƚĂŶĐĞ
ŽŵŵƵŶŝƚLJ
ŽŵŵƵŶŝƚLJ
ŽŵŵƵŶŝƚLJ
2YHUOD
1HWZRUN
/DHU
ŝƌĞĐƚŽƌLJ
ŽĐĂůƐƚĂƚĞ
ƚĂďůĞƐ
8QGHUOLQJ
1HWZRUN
/DHU
EŽĚĞƐ/ŶƚĞƌĐŽŶŶĞĐƚŝǀŝƚLJĂƌĐŚŝƚĞĐƚƵƌĞ
;hŶƐƚƌƵĐƚƵƌĞĚƵŶĚĞƌůLJŝŶŐWϮWĂƌĐŚŝƚĞĐƚƵƌĞͿ
WƌŽǀŝĚĞƌWĞĞƌƐůŝƐƚŝŶŐƐ
Figure 1.2 Multi-layered Composition Framework
allowing a dynamic cost-based service selection. In the following we briefly describe the
framework layers and components.
• The Business Layer models the business logic in a workflow of required activities.
The main purpose is to refine business activities in abstract services modeling func-
tionalities shared by applications in the domain of interest.
• The Application Layer represents the composite application by a graph of abstract
services denoted by Generic Business Process (GBP). A GBP is an oriented attributed
graph whose vertices represent abstract services and edges represent control sequences
indicating functional dependencies between the abstract services. Attributes are associ-
ated with the vertices and the edges in order to represent functional and non-functional
data and characteristics. Yetongnon et al. [46] discusses details about this layer and the
following ones.
• The Instances Layer contains a set of possible service compositions generated from
the GBP abstract service graph and based on the available services in the ecosystem.
This conversion of a GBP into a set of GBP instances is carried out by an instantia-
tion process in which services registered by peers are substituted for the abstract ser-
vices of the GBP. Thus, a GBP instance is a directed attributed graph whose nodes are
registered services, edges connect two services based on the functional dependencies](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-30-320.jpg)
![A Community-based Approach for Service-based Application Composition in an Ecosystem 17
made to the community directory entry related to SN5, because the functionality transfor-
mation is still implemented by peer E. Both SN4 and SN5 flag H as OFF in their respective
local state tables 1.3(a) and 1.3(b). SN4 and SN5 notify the remaining super-peers sending
them incremental updates of the community directory.
Table 1.2 Community Directory in the overlay based on regional locality
Super-peer Supported Abstract Services
SN1 {production, regulation, transformation, delivery}
SN2 {production, regulation, transformation, delivery}
SN3 {production, regulation, delivery}
SN4 {production, transformation, delivery}
SN5 {transformation}
Consider the following user requirements where the GBP in Figure 1.5 is instantiated by
super-peer SN4: (i) give preference to services in the community of SN4 (ii) minimize the
total execution price in terms of execution cost in $ as shown in Table 1.1. From the local
state table (Table 1.3(a)) SN4 determines the abstract services implemented by local mem-
bers, in this case we have respectively for production: (ALSTHOM-prod, B); for transfor-
mation: (DirectEnergie, C) and for delivery: (ALSTHOM-del, B). Functionalities production,
transformation and delivery are matched locally while the regulation functionality must be
matched remotely by a neighbor super-peer.
On SN4 the GBP is expressed as the union of a local and a remote subsets of abstract ser-
vices. The local subset contains abstract services implemented by the community members,
whereas the remote subset designates the set of abstract services whose implementations
will be discovered on the network. To compute the local part of the instance, SN4 asks its
underlying peers for the execution price (table 1.1) of the services matching the abstract
services in the local part of the GBP. In the chosen example each functionality is matched
by one concrete service, thus the following services are selected to compose the starting
combination: {(ALSTHOM-prod, B), ((DirectEnergie, C), (ALSTHOM-del, B)}. To compute
the remote part of the instance SN4 uses table 1.2 to find other super-peers whose commu-
nities implement the regulation abstract service. SN4 queries SN1, SN2 and SN3.
SN1 and SN3 reply respectively with (RTE, A, 0.15), (REE, D, 0.3); while SN2 replies with
both (UTCE, G, 0.25) and (RWETSO, F, 0.2). Aiming on minimizing the execution cost in $,
SN4 selects (RTE, A, 0.15). On SN4, the GBP starting instance becomes {(ALSTHOM-prod,
B), (RTE, A), (DirectEnergie, C), (ALSTHOM-del, B)}. This process is available in [46].](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-37-320.jpg)
![Bibliography 21
1.6 Conclusions and Further Research
This chapter addressed service-based application composition in ecosystems. Given an
ecosystem and a set of associated business-related properties, abstract services are de-
fined to model business-relevant functionalities in the ecosystem. Peers complying with
the ecosystem agreement provide services implementing the abstract services interfaces
and respecting their related properties.
We presented a multi-layered comprehensive framework for service-based application
composition. The associated application composition approach, first consists of modeling
the studied ecosystem’s business logic in a workflow of activities from which abstract ser-
vices are described. Second, an application is defined as a graph of abstract services. Third,
the set of application realizations is defined as a set of combinations of available services
in the ecosystem capable of executing the application process.
In order to improve peers communication effectiveness, we proposed to organize them
into communities in a hybrid overlay network based on selected properties relevant to the
ecosystem. This network organization helps us to abridge the discovery process and to
focus more on the instantiation and the failure recovery activities. We described commu-
nication related to the instantiation process in a case study of the European Electricity
Ecosystem. We presented a comparison of a simplified application instantiation example
in two different views of the ecosystem, namely, the regional locality view and the pro-
vided functionalities view. The failure recovery process and the use of other non-functional
properties as trust to build peer-communities will be addressed in future work.
Bibliography
[1] I. Vessey and A. P. Sravanapudi, Case tools as collaborative support technologies, Commun.
ACM. 38(1), 83–95, (1995). ISSN 0001-0782. doi: http://doi.acm.org/10.1145/204865.
204882.
[2] World Wide Web Consortium (W3C). Web Services Description Language (WSDL) Version
2.0. http://www.w3.org/TR/wsdl20/.
[3] A. A. Patil, S. A. Oundhakar, A. P. Sheth, and K. Verma. Meteor-s web service annotation
framework. In WWW ’04: Proceedings of the 13th international conference on World
Wide Web, pp. 553–562, New York, NY, USA, (2004). ACM. ISBN 1-58113-844-X.
doi: http://doi.acm.org/10.1145/988672.988747.
[4] D. Martin, M. Burstein, D. Mcdermott, S. Mcilraith, M. Paolucci, K. Sycara, D. L. Mcguin-
ness, E. Sirin, and N. Srinivasan, Bringing semantics to web services with owl-s, World
Wide Web. 10(3), 243–277, (2007). ISSN 1386-145X. doi: http://dx.doi.org/10.1007/
s11280-007-0033-x.
[5] D. Roman, U. Keller, H. Lausen, J. de Bruijn, R. Lara, M. Stollberg, A. Polleres, C. Feier,](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-41-320.jpg)
![22 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon
C. Bussler, and D. Fensel, Web service modeling ontology, Applied Ontology. 1(1), 77–
106, (2005).
[6] OASIS. Reference model for service oriented architecture 1.0 (October, 2006).
http://docs.oasis-open.org/soa-rm/v1.0/soa-rm.pdf.
[7] M. J. Carey, SOA what?, IEEE Computer. 41(3), 92–94, (2008).
[8] U. Keller, R. Lara, A. Polleres, I. Toma, M. Kifer, and D. Fensel. Wsmo web service discovery
(November, 2004). URL http://www.wsmo.org/2004/d5/d5.1/v0.1/20041112/.
[9] D. Booth, H. Haas, F. McCabe, E. Newcomer, M. Champion, C. Ferris, and D. Or-
chard. Web services architecture, (2004). URL http://www.w3.org/TR/2004/
NOTE-ws-arch-20040211/.
[10] I. Toma, D. Fensel, M. Moran, K. Iqbal, T. Strang, and D. Roman. An Evaluation of Discovery
approaches in Grid and Web services Environments. In The 2nd International Confer-
ence on Grid Service Engineering and Management, Erfurt, Germany (September, 2005).
URL http://www.ict.swin.edu.au/conferences/gsem2005/.
[11] T. Bellwood, S. Capell, L. Clement, J. Colgrave, M. J. D. D. Feygin, A. Hately, R. Kochman,
P. Macias, M. Novotny, M. Paolucci, C. von Riegen, T. Rogers, K. Sycara, P. Wenzel, and
Z. Wu. Uddi version 3.0.2 (October, 2004). URL http://uddi.org/pubs/uddi-v3.
0.2-20041019.htm.
[12] T. Bellwood, L. Clment, D. Ehnebuske, A. H. M. Hondo, Y. L. Husband, K. Januszewski, S. Lee,
B. McKee, J. Munter, and C. von Riegen. Uddi version 3.0 (July, 2002). URL http:
//www.uddi.org/pubs/uddi-v3.00-published-20020719.htm.
[13] T. Bellwood, D. Bryan, V. Draluk, D. Ehnebuske, T. Glover, A. Hately, Y. L. Husband, A. Karp,
K. Kibakura, C. Kurt, J. Lancelle, S. Lee, S. MacRoibeaird, A. T. Manes, B. McKee,
J. Munter, T. Nordan, C. Reeves, D. Rogers, C. Tomlinson, C. Tosun, C. von Riegen,
and P. Yendluri. Uddi version 2.04 (July, 2002). URL http://uddi.org/pubs/
ProgrammersAPI-V2.04-Published-20020719.htm.
[14] N. Srinivasan, M. Paolucci, and K. Sycara. Adding owl-s to uddi, implementation and through-
put service. In First International Workshop on Semantic Web Services and Web Process
Composition (SWSWPC 2004) 6-9, 2004, San Diego, California, USA, (2004).
[15] P. Rompothong and T. Senivongse. A query federation of uddi registries. In ISICT ’03: Proceed-
ings of the 1st international symposium on Information and communication technologies,
pp. 561–566. Trinity College Dublin, (2003).
[16] K. Sivashanmugam, K. Verma, and A. Sheth. Discovery of web services in a federated registry
environment. In ICWS ’04: Proceedings of the IEEE International Conference on Web
Services, p. 270, Washington, DC, USA, (2004). IEEE Computer Society. ISBN 0-7695-
2167-3.
[17] K. Verma, K. Sivashanmugam, A. Sheth, A. Patil, S. Oundhakar, and J. Miller, Meteor-s wsdi:
A scalable p2p infrastructure of registries for semantic publication and discovery of web
services, Information Technology and Management. 6(1), 17–39 (January, 2005). URL
http://dx.doi.org/10.1007/s10799-004-7773-4.
[18] M. Paolucci, K. Sycara, T. Nishimura, and N. Srinivasan. Using daml-s for p2p discovery. In
International Conference on Web Services (ICWS03), (2003). URL http://www.cs.
cmu.edu/~{}softagents/papers/p2p_icws.pdf.
[19] C. Schmidt and M. Parashar, A peer-to-peer approach to web service discovery, World Wide
Web. 7(2), 211–229, (2004). ISSN 1386-145X. doi: http://dx.doi.org/10.1023/B:WWWJ.
0000017210.55153.3d.
[20] M. D. Ernst, R. Lencevicius, and J. H. Perkins. Detection of web service substitutability and
composability. In WS-MaTe 2006: International Workshop on Web Services — Modeling
and Testing, pp. 123–135, Palermo, Italy (June 9,, 2006).
[21] X. Dong, A. Halevy, J. Madhavan, E. Nemes, and J. Zhang. Similarity search for web services.](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-42-320.jpg)
![Bibliography 23
In VLDB ’04: Proceedings of the Thirtieth international conference on Very large data
bases, pp. 372–383. VLDB Endowment, (2004). ISBN 0-12-088469-0.
[22] M. Paolucci, T. Kawamura, T. R. Payne, and K. P. Sycara. Semantic matching of web services
capabilities. In ISWC ’02: Proceedings of the First International Semantic Web Confer-
ence on The Semantic Web, pp. 333–347, London, UK, (2002). Springer-Verlag. ISBN
3-540-43760-6.
[23] B. Benatallah, M.-S. Hacid, A. Leger, C. Rey, and F. Toumani, On automating web services
discovery, The VLDB Journal. 14(1), 84–96, (2005). ISSN 1066-8888. doi: http://dx.
doi.org/10.1007/s00778-003-0117-x.
[24] Y. Taher, D. Benslimane, M.-C. Fauvet, and Z. Maamar. Towards an Approach for Web services
Substitution. In IDEAS ’06: Proceedings of the 10th International Database Engineering
and Applications Symposium, pp. 166–173. IEEE Computer Society (Dec., 2006). ISBN
0-7695-2577-6. doi: http://dx.doi.org/10.1109/IDEAS.2006.45. URL http://liris.
cnrs.fr/publis/?id=2464.
[25] L. Bordeaux, G. Salaün, D. Berardi, and M. Mecella. When are two web services
compatible? In eds. M.-C. Shan, U. Dayal, and M. Hsu, TES, vol. 3324, Lec-
ture Notes in Computer Science, pp. 15–28. Springer, (2004). ISBN 3-540-
25049-2. URL http://springerlink.metapress.com/openurl.asp?genre=
articleamp;issn=0302-9743amp;volume=3324amp;spage=15.
[26] OASIS. Web Services Business Process Execution Language (WS-BPEL) Version 2.0. www.
oasis-open.org/committees/wsbpel/.
[27] World Wide Web Consortium (W3C). Web Service Choreography Interface. http://www.w3.
org/TR/wsci/.
[28] S. Thakkar, C. A. Knoblock, and J. L. Ambite. A view integration approach to dynamic com-
position of web services. In In Proceedings of 2003 ICAPS Workshop on Planning for
Web Services, (2003).
[29] F. Casati, S. Ilnicki, L.-j. Jin, V. Krishnamoorthy, and M.-C. Shan. Adaptive and dynamic
service composition in eflow. In CAiSE ’00: Proceedings of the 12th International Con-
ference on Advanced Information Systems Engineering, pp. 13–31, London, UK, (2000).
Springer-Verlag. ISBN 3-540-67630-9.
[30] H. Sun, X. Wang, B. Zhou1, and P. Zou. Research and implementation of dynamic web services
composition. In Advanced Parallel Processing Technologies. pringer Berlin / Heidelberg,
(2003).
[31] J. Rao and X. Su, A survey of automated web service composition methods, Semantic Web
Services and Web Process Composition. pp. 43–54, (2005).
[32] S. Dustdar and W. Schreiner, A survey on web services composition, International journal of
web and grid services. (2005).
[33] B. Yang and H. Garcia-Molina. Designing a super-peer network. In Proceedings of the 19th
International Conference on Data Engineering, March 5-8, 2003, Bangalore, India, pp.
49–60. IEEE Computer Society, (2003).
[34] M. Khambatti, K. D. Ryu, and P. Dasgupta. Peer-to-peer communities: Formation and discov-
ery. In Proceedings of International Conference on Parallel and Distributed Computing
Systems, PDCS 2002, November 4-6, 2002, Cambridge, USA, pp. 161–166, (2002).
[35] I. Stoica, R. Morris, D. Karger, M. F. Kaashoek, and H. Balakrishnan. Chord: A scalable peer-
to-peer lookup service for internet applications. In SIGCOMM ’01: Proceedings of the
2001 conference on Applications, technologies, architectures, and protocols for computer
communications, pp. 149–160, New York, NY, USA, (2001). ACM. ISBN 1-58113-411-
8. doi: http://doi.acm.org/10.1145/383059.383071.
[36] S. Ratnasamy, P. Francis, M. Handley, R. Karp, and S. Schenker. A scalable content-addressable
network. In SIGCOMM ’01: Proceedings of the 2001 conference on Applications, tech-](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-43-320.jpg)
![24 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon
nologies, architectures, and protocols for computer communications, pp. 161–172, New
York, NY, USA, (2001). ACM. ISBN 1-58113-411-8. doi: http://doi.acm.org/10.1145/
383059.383072.
[37] K. Oztoprak and G. B. Akar. Two-way/hybrid clustering architecture for peer to peer systems.
In Proceedings of the International Conference on Internet and Web Applications and
Services (ICIW 2007), May 13-19, 2007, Le Morne, Mauritius, pp. 11–16, (2007).
[38] S. Ratnasamy, M. Handley, R. M. Karp, and S. Shenker. Topologically-aware overlay con-
struction and server selection. In Proceedings of INFOCOM 2002: Twenty-First Annual
Joint Conference of the IEEE Computer and Communications Societies., pp. 1190– 1199.
IEEE Computer Society, (2002).
[39] X. Y. Zhang, Q. Zhang, Z. Zhang, G. Song, and W. Zhu, A Construction of Locality-Aware
Overlay Network: mOverlay and Its Performance, IEEE Journal on Selected Areas in
Communications. 22(1), 18–28, (2004).
[40] R. Zhang, A. R. Butt, and Y. C. Hu. Topology-Aware Peer-to-Peer On-demand Streaming.
In Proceedings of NETWORKING 2005: Networking Technologies, Services, and Pro-
tocols; Performance of Computer and Communication Networks; Mobile and Wireless
Communication Systems, 4th International IFIP-TC6 Networking Conference, Water-
loo, Canada, May 2-6, 2005, vol. 3462, Lecture Notes in Computer Science, pp. 1–14.
Springer, (2005).
[41] L. Ramaswamy, B. Gedik, and L. Liu, Connectivity based node clustering in decentralized peer-
to-peer networks, p2p. 00, 66, (2003). doi: http://doi.ieeecomputersociety.org/10.1109/
PTP.2003.1231505.
[42] W. Zheng, S. Zhang, Y. Ouyang, F. Makedon, and J. Ford. Node clustering based on link delay
in p2p networks. In SAC ’05: Proceedings of the 2005 ACM symposium on Applied
computing, pp. 744–749, New York, NY, USA, (2005). ACM. ISBN 1-58113-964-0.
doi: http://doi.acm.org/10.1145/1066677.1066845.
[43] World Wide Web Consortium (W3C). QoS for Web Services: Requirements and Possible Ap-
proaches. http://www.w3c.or.kr/kr-office/TR/2003/ws-qos/.
[44] N. Griffiths and M. Luck. Coalition formation through motivation and trust. In AAMAS ’03:
Proceedings of the second international joint conference on Autonomous agents and mul-
tiagent systems, pp. 17–24, New York, NY, USA, (2003). ACM. ISBN 1-58113-683-8.
doi: http://doi.acm.org/10.1145/860575.860579.
[45] Y. Wang and J. Vassileva. Trust-based community formation in peer-to-peer file sharing net-
works. In WI ’04: Proceedings of the 2004 IEEE/WIC/ACM International Conference on
Web Intelligence, pp. 341–348, Washington, DC, USA, (2004). IEEE Computer Society.
ISBN 0-7695-2100-2. doi: http://dx.doi.org/10.1109/WI.2004.144.
[46] K. Yetongnon, E. Abi-Lahoud, and M. Viviani. Service based application design in P2P overlay
networks. In PDCCS-2008: ISCA 21st International Conference on Parallel and Dis-
tributed Computing and Communication Systems, New Orleans, LA, USA (September,
2008).](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-44-320.jpg)
![26 Ernastuti and Ravi A. Salim
Many communication problems are special instances of the following (N, p,k1,k2-routing
problem. N packets, each with its own source and destination, must be routed such that at
most k1 packets are initially at any node, and at most k2 packets are finally at any node. The
N packets reside on p nodes. Regular topologies offer the advantage that all nodes have
a global knowledge of the network, allowing for simple routing and scheduling decisions.
Special algorithms are also interesting if many of the packets that are sent or received by a
node are the same. Of particular importance are the following basic operations:
(1) A single node broadcast involves the transfer of a message from a particular node to all
other network nodes;
(2) A single node scatter is similar to single node broadcast except that different messages
are broadcasted;
(3) A multinode broadcast involves the simultaneous single node broadcast from all net-
work nodes (there are different messages);
(4) A total exchange (also called gossiping) is similar to multinode broadcast, except that
all the packets sent are different;
(5) A single node accumulate (also called gather) is the dual operation to single node
scatter; and
(6) A multinode accumulate is the dual operation to multinode broadcasting.
Recently the hypercube has become a popular interconnection topology for parallel and dis-
tributed processing. The popularity of the hypercube is due to its appealing properties such
as logarithmic diameter and high bisection width, ease to embed other common structures,
and many known efficient data communication schemes. Square and Palais in 1963 pro-
posed a message passing multiprocessor computer with 2k
processing nodes, in which each
node is placed at the vertex of a k-dimensional hypercube and the edges of the hypercube
are links between the processors.
A problem with the hypercube topology is that the number of nodes in a system must be
a power of 2. In practical terms, this is a severe restriction on the sizes of systems that
can be built. This restriction can be overcome by using an incomplete hypercube, i.e., a
hypercube missing certain of its nodes [6]. Unlike hypercubes, incomplete hypercubes can
be constructed with any number of nodes.
Incomplete hypercube network models which possess almost all attractive features of the
hypercube were introduced by Wu [11], Hsu [5], Munarini [9] and Ernastuti [2] in 1993,
1997, 2002 and 2007, respectively namely Fibonacci Cube (FC), extended Fibonacci cube](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-46-320.jpg)
![Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 27
(EFC), Lucas Cube (LC), extended Lucas cube (ELC). These models are the induced sub-
graphs of the hypercube that use about 1/5 fewer links than the comparable hypercube and
its size does not increase as fast as the hypercube. They are restricted to be of certain sizes,
i.e., they are Fibonacci numbers. Therefore, Sandi Klavzar [7] classified them as members
of the Fibonacci cubes family. Though there are more Fibonacci numbers than numbers
being power of 2, they do not fill the gap left by hypercubes very well. It can be shown that
the node degrees in the FC, EFC, LC and ELC is a logarithmic function of the total number
of nodes. This property provides improved fault tolerance over the incomplete hypercubes.
They can be viewed as hypercubes with faulty nodes. They provides more choices of net-
work size to the family of cube based structures. It has been also shown in [5, 11, 9, 4]
that the FC, EFC, LC and ELC can be efficiently embedded many interesting structures
such as hypercubes, linear arrays, rings and meshes. All these make them to be attractive
interconnection topologies.
The FC, EFC, LC and ELC have similar properties except in Hamiltonicity property
[5, 11, 9, 3]. FC has Hamiltonian paths for every n, but only less than a third of them
has Hamiltonian cycles. EFC and ELC have both Hamiltonian paths and Hamiltonian cy-
cles for every n. As for LC, it has no Hamiltonian cycles at all for every n, albeit still
has Hamiltonian paths in some n’s. The incomplete hypercube can be viewed as resulting
from a complete hypercube after some nodes become faulty and the system is reconfigured
[1]. Therefore, the FC, EFC, LC and ELC not only allow the construction of systems of
arbitrary sizes, but also expose the nature of hypercube systems operating in a gracefully
degraded mode. The incomplete hypercube with N nodes, where N could be any positive in-
teger, is constructed in the same way as the hypercube. In other words, nodes are numbered
from 0 to N − 1 and two nodes are linked if and only if their binary representations differ
in exactly one bit. The incomplete hypercube suffers from a low degree of fault tolerance
under certain condition.
The reliability of data processing and data communication is very important in hypercube
systems as in all parallel systems [1]. Efficient routing and broadcasting messages is a key
issue to the performance of parallel and/or distributed systems. An important property of
any message routing algorithm is to avoid deadlock [6]. The speed and the tolerance may
be decreased if one or more processors or links become faulty [1]. In order to determine
and avoid the faulty nodes and links in the data communication, there are many different
kinds of methods to find the shortest paths between the source and the target nodes. In fact
FC, EFC and LC have been proved possessing a simple routing algorithm [5, 11, 9].](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-47-320.jpg)
![28 Ernastuti and Ravi A. Salim
In this paper, the focus is to study the data communication aspects in the extended Lucas
cubes (ELC). We use the basic operations of communication model for problem of a single
node broadcast which involves the transfer of a message from a particular node to all other
network nodes. Problem of a single node broadcast consists of three basic types of data
routing, i.e., one-to-one (unicast), one-to-all (broadcast) and one-to-many (multicast). In
this paper we apply data routing algorithms for ELC which refer to [10]. We show the
unicast algorithm which uses a Hamming distance path for any two nodes for ELC. We also
show the broadcast algorithm which employs the extended Lucas tree, and then we present
two heuristic multicast algorithms based on an extended Lucas tree and a Hamiltonian cycle
on ELC, respectively. To measure the efficiency of routing algorithms, the time and traffic
steps are used.
2.2 Preliminaries and Notations
We represent an interconnection topology by a graph G = (V,E), whereV (the set of nodes)
denotes the processors and E (the set of edges) represents the communication links between
processors; an edge is an unordered pair xy = {x,y} of distinct nodes of G. Sometimes, to
avoid ambiguity, V and E are denoted by VG and EG. And we denote the number of nodes
and edges of G by |VG| and |EG|.
Definition 2.1. A path on a graph (also called a chain) is a sequence x1,x2,...,xn such that
{x1,x2}, {x2,x3},...,{xn−1,xn}, are edges of the graph and the xi are distinct. A closed
path (x1,x2,...,xn,x1) on a graph is called a graph cycle or circuit.
Definition 2.2. For x, y ∈ VG, dG(x,y) or d(x,y), denotes the length of a shortest path (a
path with the least number of edges) in G from x to y.
Let {0, 1}n denote the set of length n binary strings.
Definition 2.3. The Hamming distance between two binary strings x, y ∈ {0, 1}n denoted
H(x,y), is the number of bits where x and y differ.
Definition 2.4. The Hypercube of dimension n, denoted by Q(n), is the graph, where the
set of labels of nodes is {0, 1}n
and two nodes x and y are adjacent if and only if their labels
differ in exactly one bit (in other words H(x,y) = 1).
Fig. 2.1 shows examples of Q(n), for n = 1, 2, 3, 4 respectively.](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-48-320.jpg)
![Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 29
Figure 2.1 Hypercube of dimension 1, 2, 3, 4: Q(1), Q(2), Q(3) and Q(4)
Definition 2.5. The Hamming distance between two binary strings x, y ∈ {0, 1}n denoted
H(x,y) shows the length of shortest path between node x and node y.
Definition 2.6. The Fibonacci numbers form a sequence of positive integers fn, where
f1 = 1, f2 = 1 and fn = fn−1 + fn−2, for n 2.
Definition 2.7. A Fibonacci string of length n is a binary string a1a2 ...an which belongs
to {0, 1}n with aiai+1 = 0, 1 ⩽ i n. In other words, a Fibonacci string is a binary string
of length n with no two consecutive ones.
It is easy to see that the number of Fibonacci strings of length n is the (n + 2) Fibonacci
number (this connects Definition 2.6 and 2.7). The definition of FC, EFC, LC and ELC are
based upon Fibonacci strings and the Hamming distance.
2.3 Graph Models of Fibonacci Cube Family
FC, EFC and LC topologies use the Fibonacci sequence; however the initial conditions
among them may differ from the initial conditions of the Fibonacci sequence. In this section
we show the differences. The symbol · denotes a concatenation operation; for example,
01 · {0,1} = {010,011} and 01 · { } = {01}. The FC, EFC and LC can be respectively
described as below.
Definition 2.8 (Fibonacci cube [5]). For n ⩾ 0, the Fibonacci cube FC(n) =
(VFC(n),EFC(n)) is defined as follows:
VFC(n), the set of labels of nodes in FC(n), is recursively defined as
VFC(n) =
⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
∅ if n = 0
{λ} if n = 1, 2
{0, 1} if n = 3
0 ·VFC(n −1)∪10 ·VFC(n −2) if n 3](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-49-320.jpg)
![30 Ernastuti and Ravi A. Salim
Two nodes in VFC(n) are connected by an edge in EFC(n) if and only if their labels differ
exactly in one position.
An FC(n) contains two disjoint subgraphs that are isomorphic to FC(n−1) and FC(n−2)
[5]. Fig. 2.2 shows examples of FC(n), with n = 3, 4, 5, 6 respectively.
Property 2.1 ([5]). For any n ⩾ 3, |VFC(n) = fn, where fn is the nth Fibonacci number.
Figure 2.2 Fibonacci cube (a) FC(3), (b) FC(4), (c) FC(5), (d) FC(6)
Definition 2.9 (Extended Fibonacci cube [11]). For n ⩾ 0, the extended Fibonacci cube
EFC(n) = (VEFC(n),EEFC(n)) is defined as follows:
VEFC(n), the set of labels of nodes in EFC(n), is recursively defined as
VEFC(n) =
⎧
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩
∅ if n = 0
{λ} if n = 1, 2
{0, 1} if n = 3
{00, 10, 11, 01} if n = 4
0 ·VEFC(n −1)∪10 ·VEFC(n −2) if n 4
Two nodes in VEFC(n) are connected by an edge in EEFC(n) if and only if their labels differ
exactly in one position.
An EFC(n) contains two disjoint subgraphs that are isomorphic to EFC(n − 1) and
EFC(n −2) [11]. Fig. 2.3 shows examples of EFC(n), with n = 3, 4, 5, 6 respectively.
Property 2.2 ([7]). The number of nodes of EFC(n) is 2 fn−1, where fn is the nth Fibonacci
number.
Definition 2.10. An extended Fibonacci tree T1(n) of EFC(n) is defined as follows: (Base)
T1(3) and T1(4) are defined as shown in Fig. 2.4a and 2.4b. Basically, T1(3) is EFC(3) with
node 0 being the root and T1(4) is an EFC(4) rooted at node 00 after removing the link](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-50-320.jpg)
![Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 31
Figure 2.3 Extended Fibonacci cube (a) EFC(3), (b) EFC(4), (c) EFC(5), (d) EFC(6)
connecting nodes 01 and 11. (Recursion) T1(n) (n 4) consists of T1(n −1) and T1(n −2)
by connecting the root of T1(n −2) as a child of the root of T1(n −1). Suppose T1(n) also
denotes the set of nodes in T1(n), then T1(n) = 0 ·T1(n −1)∪10 ·T1(n −2).
Fig. 2.4 shows examples of the extended Fibonacci tree T1(n), for n = 3, 4, 5, 6 respectively.
Figure 2.4 Extended Fibonacci tree (a) T1(3), (b) T1(4), (c) T1(5), (d) T1(6)
Definition 2.11 (Lucas cube [9]). For n ⩾ 0, the Lucas cube LC(n) = (VLC(n),ELC(n)) is
defined as follows:
VLC(n), the set of labels of nodes in LC(n), is recursively defined as
VLC(n) =
⎧
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎩
∅ if n = 0
{λ} if n = 1, 2
{0, 1} if n = 3
0 ·VFC(n −1)∪10 ·VFC(n −3)·0 if n 3
Two nodes in VLC(n) are connected by an edge in ELC(n) if and only if their labels differ
exactly in one position.
For any n ⩾ 0, The FC(n), EFC(n) and LC(n) are induced subgraphs of Q(n−2) [5, 11, 9].](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-51-320.jpg)
![32 Ernastuti and Ravi A. Salim
2.4 Extended Lucas Cube (ELC)
The ELC is defined on the same way of LC recurrence by using the EFC as its initial
condition. The following definition gives a recursive definition for ELC.
Definition 2.12 (Extended Lucas cube [3, 2]). For n ⩾ 0, the extended Lucas cube
ELC(n) = (VELC(n),EELC(n)) is defined as follows:
VELC(n), the set of labels of nodes in ELC(n), is recursively defined as
VELC(n) =
⎧
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩
∅ if n = 0
{λ} if n = 1, 2
{0, 1} if n = 3
{00, 10, 11, 01} if n = 4
0 ·VEFC(n −1)∪10 ·VEFC(n −3)·0 if n 4
Two nodes in VELC(n) are connected by an edge in EELC(n) if and only if their labels differ
exactly in one position.
Definition 2.12 says thatVELC(n) ⊆ {0, 1}n−2, n ⩾ 2, and so ELC(n) is an induced subgraph
of Q(n −2). Fig. 2.5 shows examples of ELC(n), with n = 3, 4, 5, 6 respectively.
Figure 2.5 Extended Lucas cube (a) ELC(3), (b) ELC(4), (c) ELC(5), (d) ELC(6)
The properties of the extended Lucas cube (ELC) are given below. Refer to [2] for more
detail and proofs to these properties.
Property 2.3. For n ⩾ 3, ELC(n) contains two disjoint subgraphs that are isomorphic to
ELC(n −1) and ELC(n −3), respectively.
Property 2.4. For n ⩾ 4, n = 5, ELC(n) is a Hamiltonian graph.
Property 2.5. There exists a Hamming distance path between any two nodes in ELC.](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-52-320.jpg)
![Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 33
Property 2.6. The Hamming distance path between any two nodes in ELC is a shortest
path.
[2] has proved that ELC(n) contains two subgraphs which are isomorphic to EFC(n − 1)
and EFC(n −3), respectively. There are exactly fn−3 edges linking those two subgraphs.
Property 2.7. For any n ⩾ 5, in ELC(n), there are exactly fn−3 edges linking subgraph
induced by 0 ·VEFC(n −1) to subgraph induced by 10 ·VEFC(n −1)·0.
Property 2.8. For any n ⩾ 3, ELC(n) is a connected graph.
Property 2.9. Diameter of ELC(n) is n −2, for n ⩾ 3.
Property 2.10. The node degree of a node in ELC(n), n ⩾ 3, is between
n −3
3
and n−2,
except for n = 4, the node degree is n −2.
Property 2.11. For n ⩾ 3, |VELC(n)| = |VEFC(n −1)|+|VEFC(n −3)|.
Property 2.12. For any n ⩾ 5, the number of nodes of ELC(n) is 2 fn−2 +2 fn−4.
Table 2.1 shows the number of nodes of hypercube, FC(n), EFC(n), LC(n) and ELC(n) for
3 ⩽ n ⩽ 12.
Definition 2.13. An extended Lucas tree T2(n) of ELC(n) is defined as follows: (Base)
T2(3) and T2(4) are defined as shown in Fig. 2.6a and 2.6b. Basically, T2(3) is ELC(3)
with node 0 being the root and T2(4) is an ELC(4) rooted at node 00. (Recursion) T2(n)
(n ⩾ 4) consists of T1(n − 1) and T1(n − 3) by connecting the root of T1(n − 3) as a child
of the root of T1(n−1). Suppose T2(n) also denotes the set of nodes in T2(n), then T2(n) =
0 ·T1(n −1)∪10 ·T1(n −3)·0.
Property 2.13. Extended Lucas Tree T2(n) is a spanning tree of T2(n).
Property 2.14. T2(n) contains two disjoint subtrees that are isomorphic to T1(n − 1) and
T1(n −3), respectively.
Property 2.15. For any n ⩾ 3, the span of T2(n) is n −2.
Property 2.16. For any n ⩾ 3, the height of T2(n) is
n −2
2
.
Fig. 2.6 shows examples of the extended Lucas tree T2(n), for n = 1, 2, 3, 4 respectively.
Property 2.17. In T2(n), the children of the root are dimension ordered, i.e. the ith child of
the root is the neighbor of the root on the ith dimension.](https://image.slidesharecdn.com/2120406-250519152245-e845e5b3/85/Webbased-Information-Technologies-And-Distributed-Systems-Alban-Gabillon-53-320.jpg)
Webbased Information Technologies And Distributed Systems Alban Gabillon
- 1. Webbased Information Technologies And Distributed Systems Alban Gabillon download https://ebookbell.com/product/webbased-information-technologies- and-distributed-systems-alban-gabillon-4240812 Explore and download more ebooks at ebookbell.com
- 2. Here are some recommended products that we believe you will be interested in. You can click the link to download. Solutions And Innovations In Webbased Technologies For Augmented Learning Improved Platforms Tools And Applications Advances In Webbased Learning 1st Edition Nikos Karacapilidis https://ebookbell.com/product/solutions-and-innovations-in-webbased- technologies-for-augmented-learning-improved-platforms-tools-and- applications-advances-in-webbased-learning-1st-edition-nikos- karacapilidis-2525152 Accessibility Of Webbased Information Resources For People With Disabilities Part 1 Axel Schmetzke https://ebookbell.com/product/accessibility-of-webbased-information- resources-for-people-with-disabilities-part-1-axel-schmetzke-5070748 Web Based Energy Information And Control Systems Case Studies And Applications B L Capehart https://ebookbell.com/product/web-based-energy-information-and- control-systems-case-studies-and-applications-b-l-capehart-987360 Handbook Of Web Based Energy Information And Control Systems Barney L Capehart https://ebookbell.com/product/handbook-of-web-based-energy- information-and-control-systems-barney-l-capehart-2392600
- 3. Visual Information Systems Webbased Visual Information Search And Management 10th International Conference Visual 2008 Salerno Italy September 1112 2008 Proceedings 1st Edition Gennady Andrienko https://ebookbell.com/product/visual-information-systems-webbased- visual-information-search-and-management-10th-international- conference-visual-2008-salerno-italy- september-1112-2008-proceedings-1st-edition-gennady-andrienko-2040210 Creating Webbased Laboratories Advanced Information And Knowledge Processing 1st Edition Cc Ko https://ebookbell.com/product/creating-webbased-laboratories-advanced- information-and-knowledge-processing-1st-edition-cc-ko-2156900 Webbased Learning Through Educational Informatics Information Science Meets Educational Computing 1st Edition Nigel Ford https://ebookbell.com/product/webbased-learning-through-educational- informatics-information-science-meets-educational-computing-1st- edition-nigel-ford-2177296 Webbased Education Concepts Methodologies Tools And Applications 1st Edition Information Resources Management Association https://ebookbell.com/product/webbased-education-concepts- methodologies-tools-and-applications-1st-edition-information- resources-management-association-1951222 Knowledgebased Information Retrieval And Filtering From The Web 1st Edition Krzysztof Wcel Auth https://ebookbell.com/product/knowledgebased-information-retrieval- and-filtering-from-the-web-1st-edition-krzysztof-wcel-auth-4200116
- 5. ATLANTIS AMBIENT AND PERVASIVE INTELLIGENCE VOLUME 2 SERIES EDITOR: ISMAIL KHALIL
- 6. Atlantis Ambient and Pervasive Intelligence Series Editor: Ismail Khalil, Linz, Austria (ISSN: 1875-7669) Aims and scope of the series The book series ‘Atlantis Ambient and Pervasive Intelligence’ publishes high quality titles in the fields of Pervasive Computing, Mixed Reality, Wearable Computing, Location-Aware Computing, Ambient Interfaces, Tangible Interfaces, Smart Environments,Intelligent Inter- faces, Software Agents and other related fields. We welcome submission of book proposals from researchers worldwide who aim at sharing their results in this important research area. All books in this series are co-published with World Scientific. For more information on this series and our other book series, please visit our website at: www.atlantis-press.com/publications/books AMSTERDAM – PARIS c ATLANTIS PRESS / WORLD SCIENTIFIC
- 7. Web-Based Information Technologies and Distributed Systems Alban Gabillon University of Polynésie Française BP 6570 98702 FAA’A Tahiti Polynésie française Quan Z. Sheng School of Computer Science University of Adelaide Adelaide, SA 5005 Australia Wathiq Mansoor American University in Dubai, UAE AMSTERDAM – PARIS
- 8. Atlantis Press 29, avenue Laumière 75019 Paris, France For information on all Atlantis Press publications, visit our website at: www.atlantis-press.com Copyright This book, or any parts thereof, may not be reproduced for commercial purposes in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system known or to be invented, without prior permission from the Publisher. Atlantis Ambient and Pervasive Intelligence Volume 1: Agent-Based Ubiquitous Computing - Eleni Mangina, Javier Carbo, José M. Molina ISBN: 978-90-78677-28-4 ISSN: 1875-7669 c 2010 ATLANTIS PRESS / WORLD SCIENTIFIC e-ISBN: 978-94-91216-32-9
- 9. Preface The Fourth International Conference on Signal-Image Technology Internet-Based Sys- tems (SITIS 2008) has been successfully held during the period 30th November to 3rd of December of the year 2008 in Bali, Indonesia. The Track Web-Based Information Tech- nologies Distributed Systems (WITDS) is one of the four tracks of the conference. The track is devoted to emerging and novel concepts, architectures and methodologies for cre- ating an interconnected world in which information can be exchanged easily, tasks can be processed collaboratively, and communities of users with similar interests can be formed while addressing security threats that are present more than ever before. The track has attracted a large number of submissions; only fifteen papers have been accepted with ac- ceptance rate 27 %. After the successful presentations of the papers during the conference, the track chairs have agreed with Atlantis publisher to publish the extended versions of the papers in a book. Each paper has been extended with a minimum of 30 % new materials from its original conference manuscript. This book contains these extended versions as chapters after a second round of reviews and improvement. The book is an excellent resource of information to researchers and it is based on four themes; the first theme is on advances in ad-hoc and routing protocols, the second theme focuses on the latest techniques and methods on intelligent systems, the third theme is a latest trend in Security and Policies, and the last theme is applications of algorithms design methodologies on web based systems. We would like to give our great appreciations to the authors and the PC members of the track to their excellent contributions and effort that makes the creation of this book is achievable. Also, we would like to thank Atlantis publisher who has agreed to publish this v
- 10. vi Web-Based Information Technologies and Distributed Systems valuable book to the community. Special thanks to Zeger Karssen and Zakaria Maamar for their help and support during the publication of the book. Alban Gabillon (University of Polynésie Française, France) Quan Z. Sheng (University of Adelaide, Australia) Wathiq Mansoor (American University in Dubai, UAE)
- 11. Contents Preface v 1. A Community-based Approach for Service-based Application Composition in an Ecosystem 1 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1.1 Objectives and Contributions . . . . . . . . . . . . . . . . . . . 2 1.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.1 Service Orientation . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.2 P2P Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.3 A Framework for Sharing Services . . . . . . . . . . . . . . . . . . . . . 7 1.3.1 Ecosystem, Peer-communities and Services . . . . . . . . . . . 7 1.3.2 Multi-layered Service-based Composition Framework . . . . . . 9 1.4 The Overlay Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.4.1 Overlay Organization . . . . . . . . . . . . . . . . . . . . . . . 11 1.4.2 Super-peers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.4.3 Event Related Communication . . . . . . . . . . . . . . . . . . 13 1.5 Case Study: The European Electricity Market . . . . . . . . . . . . . . . 13 1.5.1 A Regional Locality-based Overlay . . . . . . . . . . . . . . . . 15 1.5.2 A Functionality-based Overlay . . . . . . . . . . . . . . . . . . 18 1.5.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 1.6 Conclusions and Further Research . . . . . . . . . . . . . . . . . . . . . 21 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 vii
- 12. viii Web-Based Information Technologies and Distributed Systems 2. Complexity Analysis of Data Routing Algorithms in Extended Lucas Cube Networks 25 Ernastuti and Ravi A. Salim 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.2 Preliminaries and Notations . . . . . . . . . . . . . . . . . . . . . . . . 28 2.3 Graph Models of Fibonacci Cube Family . . . . . . . . . . . . . . . . . 29 2.4 Extended Lucas Cube (ELC) . . . . . . . . . . . . . . . . . . . . . . . . 32 2.5 Data Routing Algorithms in ELC . . . . . . . . . . . . . . . . . . . . . . 34 2.5.1 Unicast (One-to-one) . . . . . . . . . . . . . . . . . . . . . . . 35 2.5.2 Broadcast (one-to-all) . . . . . . . . . . . . . . . . . . . . . . . 37 2.5.3 Multicast (One-to-many) . . . . . . . . . . . . . . . . . . . . . 40 2.5.4 Conclusion and Remark . . . . . . . . . . . . . . . . . . . . . . 41 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 3. An Incremental Algorithm for Clustering Search Results 43 Y. Liu, Y. Ouyang, H. Sheng, Z. Xiong 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 3.2 Similarity Measure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3.2.1 Similarity Measure . . . . . . . . . . . . . . . . . . . . . . . . 45 3.2.2 Document Similarity Measure . . . . . . . . . . . . . . . . . . . 47 3.3 Document Clustering . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 3.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.4.1 Test Data and Experiment . . . . . . . . . . . . . . . . . . . . . 49 3.4.2 Evaluation Measures . . . . . . . . . . . . . . . . . . . . . . . . 50 3.4.3 Evaluation of ICA . . . . . . . . . . . . . . . . . . . . . . . . . 52 3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4. Query Planning in DHT Based RDF Stores 57 D. Battré
- 13. Contents ix 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 4.2 Related work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.3 Foundation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4.4 Query Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.4.1 Selection of lookups (triple pattern and lookup position) . . . . . 65 4.4.2 Local heuristics . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.4.3 Network heuristics . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.4.4 Wrappers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.4.5 Network Heuristics (cont.) . . . . . . . . . . . . . . . . . . . . 74 4.4.6 Processing Triple Patterns . . . . . . . . . . . . . . . . . . . . . 75 4.5 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 4.5.1 Network Heuristics . . . . . . . . . . . . . . . . . . . . . . . . 83 4.6 Conclusion and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 5. A Formal Methodology to Specify Hierarchical Agent-Based Systems 93 C. Molinero, C. Andrés, and M. Núñez 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.2 Overview of some relevant articles in the field of “agents” . . . . . . . . 97 5.2.1 Pattie Maes - The dynamics of action selection . . . . . . . . . . 97 5.2.2 Yoav Shoham - Agent-oriented programming . . . . . . . . . . . 98 5.2.3 Rodney A. Brooks - Elephants don’t play chess . . . . . . . . . 100 5.3 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 5.4 Definition of the formalism . . . . . . . . . . . . . . . . . . . . . . . . . 103 5.5 The A tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 5.6 Conclusions and future work . . . . . . . . . . . . . . . . . . . . . . . . 111 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 6. Reducing Redundant Web Crawling Using URL Signatures 115 L.-K. Soon and S.H. Lee 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
- 14. x Web-Based Information Technologies and Distributed Systems 6.2 Web Crawling and the Standard URL Normalization . . . . . . . . . . . 118 6.2.1 Web Crawling . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 6.2.2 The Standard URL Normalization . . . . . . . . . . . . . . . . . 120 6.3 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 6.4 URL Signatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 6.4.1 Metadata Considered . . . . . . . . . . . . . . . . . . . . . . . 124 6.4.2 Definition of URL Signatures . . . . . . . . . . . . . . . . . . . 126 6.4.3 Application of URL Signatures . . . . . . . . . . . . . . . . . . 127 6.5 Experiments and Evaluation Metrics . . . . . . . . . . . . . . . . . . . . 129 6.5.1 Experimental Dataset . . . . . . . . . . . . . . . . . . . . . . . 129 6.5.2 Process Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 6.5.3 Evaluation Metrics . . . . . . . . . . . . . . . . . . . . . . . . . 132 6.6 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . 133 6.6.1 Experimental Results and Findings . . . . . . . . . . . . . . . . 133 6.6.2 Comparative Study with Other Methods . . . . . . . . . . . . . 135 6.6.3 Limitation of URL Signatures . . . . . . . . . . . . . . . . . . . 138 6.7 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . . 138 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 7. Interoperability Among Heterogeneous Systems in Smart Home Environment 141 T. Perumal, A.R. Ramli, C.Y. Leong, K. Samsudin, and S. Mansor 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 7.2 Background and Related Work . . . . . . . . . . . . . . . . . . . . . . . 143 7.2.1 Common Object Request Broker Architecture (CORBA) . . . . 145 7.2.2 Component Object Model (COM) . . . . . . . . . . . . . . . . . 145 7.2.3 Microsoft .NET Framework . . . . . . . . . . . . . . . . . . . . 146 7.2.4 Java Middleware Technologies . . . . . . . . . . . . . . . . . . 147 7.2.5 Web Services . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 7.3 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 7.3.1 System Architecture . . . . . . . . . . . . . . . . . . . . . . . . 149 7.3.2 Home Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 7.3.3 Database module . . . . . . . . . . . . . . . . . . . . . . . . . 152
- 15. Contents xi 7.4 System Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 7.4.1 System Elements . . . . . . . . . . . . . . . . . . . . . . . . . 153 7.4.2 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . . 153 7.5 Conclusion and Outlooks . . . . . . . . . . . . . . . . . . . . . . . . . . 155 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 8. A Formal Framework to Specify and Deploy Reaction Policies 159 F. Cuppens, N. Cuppens-Boulahia, W. Kanoun, and A. Croissant 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 8.2 Attack Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 8.2.1 LAMBDA Language and Semi-Explicit Correlation . . . . . . . 162 8.2.2 Recognizing Intrusion Objectives . . . . . . . . . . . . . . . . . 164 8.3 Countermeasure Modeling . . . . . . . . . . . . . . . . . . . . . . . . . 165 8.4 Reaction policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 8.4.1 The OrBAC Model . . . . . . . . . . . . . . . . . . . . . . . . 167 8.4.2 Using OrBAC to Specify Reaction Policy . . . . . . . . . . . . . 168 8.4.3 Security Requirements Interpretation . . . . . . . . . . . . . . . 170 8.4.4 Strategies to Manage Conflicts . . . . . . . . . . . . . . . . . . 172 8.5 Deployment of the Reaction Workflow . . . . . . . . . . . . . . . . . . . 173 8.6 Reaction Workflow Architecture . . . . . . . . . . . . . . . . . . . . . . 178 8.6.1 Low Level Reaction . . . . . . . . . . . . . . . . . . . . . . . . 178 8.6.2 Intermediate Level Reaction . . . . . . . . . . . . . . . . . . . . 179 8.6.3 High Level Reaction . . . . . . . . . . . . . . . . . . . . . . . . 180 8.7 VoIP Use Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 8.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 9. A new distributed IDS based on CVSS framework 189 J. Aussibal and L. Gallon 9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 9.2 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
- 16. xii Web-Based Information Technologies and Distributed Systems 9.3 Alert scoring tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 9.3.1 CVE Dictionary . . . . . . . . . . . . . . . . . . . . . . . . . . 194 9.3.2 CVSS Framework . . . . . . . . . . . . . . . . . . . . . . . . . 194 9.4 Our proposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 9.4.1 General principles . . . . . . . . . . . . . . . . . . . . . . . . . 200 9.4.2 Detection entity . . . . . . . . . . . . . . . . . . . . . . . . . . 201 9.4.3 Heterogeneity of local probes . . . . . . . . . . . . . . . . . . . 203 9.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 10. Modeling and Testing Secure Web Applications 207 W. Mallouli, M. Lallali, A. Mammar, G. Morales, and A.R. Cavalli 10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 10.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 10.3 Testing Methodology Overview . . . . . . . . . . . . . . . . . . . . . . 211 10.4 Functional Specification of Web Applications using IF Language . . . . . 212 10.4.1 Modeling Communicating Systems . . . . . . . . . . . . . . . . 212 10.4.2 IF Formal Language . . . . . . . . . . . . . . . . . . . . . . . . 214 10.4.3 Case Study: Travel Web Application . . . . . . . . . . . . . . . 215 10.4.4 Travel IF Specification . . . . . . . . . . . . . . . . . . . . . . 216 10.5 Secure Specification of Web Applications . . . . . . . . . . . . . . . . . 217 10.5.1 Security Rules Specification Using Nomad Language . . . . . . 217 10.5.2 Security Integration Methodology . . . . . . . . . . . . . . . . . 219 10.5.3 Correctness Proof of the Integration Approach . . . . . . . . . . 233 10.5.4 Travel Security Specification Using Nomad Language . . . . . . 235 10.5.5 Automatic Rules Integration . . . . . . . . . . . . . . . . . . . . 236 10.5.6 Rules Integration Results . . . . . . . . . . . . . . . . . . . . . 238 10.6 Test Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238 10.6.1 TestGen-IF tool . . . . . . . . . . . . . . . . . . . . . . . . . . 238 10.6.2 Fixing the Test Objectives . . . . . . . . . . . . . . . . . . . . . 241 10.6.3 Test Generation with TestGen-IF . . . . . . . . . . . . . . . . . 243 10.7 Test Cases Instantiation and Execution . . . . . . . . . . . . . . . . . . . 244 10.7.1 Tclwebtest tool . . . . . . . . . . . . . . . . . . . . . . . . . . 244
- 17. Contents xiii 10.7.2 Test Cases Instantiation . . . . . . . . . . . . . . . . . . . . . . 245 10.7.3 Test Cases Execution . . . . . . . . . . . . . . . . . . . . . . . 251 10.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 11. Secure interoperability with O2O contracts 257 C. Coma, N. Cuppens-Boulahia, and F. Cuppens 11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 11.2 Usual Approaches for Interoperability . . . . . . . . . . . . . . . . . . . 259 11.2.1 Federated Identity Management . . . . . . . . . . . . . . . . . . 259 11.2.2 Negotiation policy . . . . . . . . . . . . . . . . . . . . . . . . . 260 11.2.3 Ontological approaches . . . . . . . . . . . . . . . . . . . . . . 262 11.3 Generic Interoperation Policies . . . . . . . . . . . . . . . . . . . . . . . 264 11.3.1 Contextual Security Policy: the OrBAC model . . . . . . . . . . 264 11.3.2 Interoperability Framework: O2O principles . . . . . . . . . . . 266 11.4 Interoperability Establishment Steps: the O2O process . . . . . . . . . . 267 11.5 Interoperability Contract . . . . . . . . . . . . . . . . . . . . . . . . . . 268 11.6 Interoperability Contract Specification . . . . . . . . . . . . . . . . . . . 269 11.6.1 Underivability and Exception . . . . . . . . . . . . . . . . . . . 270 11.6.2 Compatibility Relation Patterns . . . . . . . . . . . . . . . . . . 271 11.6.3 Contract example . . . . . . . . . . . . . . . . . . . . . . . . . 273 11.7 Secure Interoperability Policy Establishment . . . . . . . . . . . . . . . 274 11.7.1 Ontological Mapping . . . . . . . . . . . . . . . . . . . . . . . 274 11.7.2 Establishment of Compatibility Relations . . . . . . . . . . . . . 276 11.8 Derivation of the Interoperability Security Policy . . . . . . . . . . . . . 277 11.8.1 Derivation rules . . . . . . . . . . . . . . . . . . . . . . . . . . 277 11.8.2 Example of derivation of an interoperability rule . . . . . . . . . 278 11.9 VPO management: Secure interoperation policy management . . . . . . . 279 11.10 AdOrBAC: interoperability policy administration . . . . . . . . . . . . . 282 11.10.1 AdOrBAC administration views . . . . . . . . . . . . . . . . . . 282 11.10.2 Licence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284 11.11 Privacy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284 11.11.1 XML-BB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
- 18. xiv Web-Based Information Technologies and Distributed Systems 11.11.2 Obfuscation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286 11.12 Illustration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 11.12.1 P2P and interoperability . . . . . . . . . . . . . . . . . . . . . . 287 11.12.2 Obfuscation during interoperability . . . . . . . . . . . . . . . . 288 11.12.3 P2P and O2O contract . . . . . . . . . . . . . . . . . . . . . . . 288 11.13 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290 12. ADMON: I/O Workload Management by Visage Administration and Monitoring Service 293 S. Traboulsi, J. Jorda, and A. M’zoughi 12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293 12.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 12.3 The Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296 12.4 ViSaGe Environment and Architecture . . . . . . . . . . . . . . . . . . . 296 12.5 Admon Functionalities and API . . . . . . . . . . . . . . . . . . . . . . 298 12.5.1 ViSaGe Monitoring . . . . . . . . . . . . . . . . . . . . . . . . 299 12.5.2 ViSaGe Administration . . . . . . . . . . . . . . . . . . . . . . 301 12.6 Admon: I/O Workload Performance . . . . . . . . . . . . . . . . . . . . 302 12.6.1 Admon Predictor Model . . . . . . . . . . . . . . . . . . . . . . 303 12.6.2 Experimental Setup and Validation with ViSaGe . . . . . . . . . 304 12.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309 13. Extracting Neglected Content from Community-type-content 311 A. Nadamoto, E. Aramaki, T. Abekawa, and Y. Murakami 13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311 13.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 13.3 Basic Concept of Content Hole . . . . . . . . . . . . . . . . . . . . . . . 315 13.4 Extracting Neglected Content . . . . . . . . . . . . . . . . . . . . . . . . 318 13.4.1 Creating a Comment Tree Structure . . . . . . . . . . . . . . . . 319
- 19. Contents xv 13.4.2 Automatic dialog corpus building . . . . . . . . . . . . . . . . . 322 13.4.3 Extracting possibly neglected content . . . . . . . . . . . . . . . 325 13.4.4 Filtering unrelated content . . . . . . . . . . . . . . . . . . . . . 326 13.4.5 Extracting neglected content . . . . . . . . . . . . . . . . . . . 326 13.4.6 Prototype System . . . . . . . . . . . . . . . . . . . . . . . . . 326 13.5 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328 13.5.1 Content Relevance and Functional Relevance . . . . . . . . . . . 328 13.5.2 Accuracy of Neglected Content . . . . . . . . . . . . . . . . . . 329 13.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
- 21. Chapter 1 A Community-based Approach for Service-based Application Composition in an Ecosystem Elie Abi-Lahoud, Marinette Savonnet, Marie-Noelle Terrasse, Marco Viviani, Kokou Yétongnon Université de Bourgogne – Sciences et Techniques, Laboratoire LE2I – Mirande, Aile de l’Ingénieur, 9, av. Savary, 21078 Dijon cedex, France The design of composite applications by combining existing services with known seman- tics is an ongoing topic in current research. Several studies are aimed at providing service description models and standards, service discovery and matching etc. However, service composition in distributed dynamic environments such as P2P ecosystems has received little attention from research communities. In this paper we present a design framework for composing services, taking in particular into account different ways of building peer- communities based on network or services characteristics. 1.1 Introduction Service oriented computing provides software designer with new concepts and emerging principles for developing loosely-coupled, cross-enterprise business applications. Tradi- tionally, software development approaches rely on CASE tools [1] and modeling concepts to describe and implement software components that can be integrated into applications. Recently, we are witnessing a shift from this static view of software development and de- ployment towards a dynamic, adaptable service based view of software design in which applications could be realized in a flexible manner to respond to changing needs of users. In this emerging design view, services provide high level functional components that can be shared in open distributed environments. The goal is to design composite applications by combining existing service components with known semantics, spanning organizations and computing platforms. Many research efforts have been aimed at service oriented computing, ranging from tech- A. Gabillon et al., Web-Based Information Technologies and Distributed Systems, 1 Atlantis Ambient and Pervasive Intelligence 2, DOI 10.2991/978-94-91216-32-9_1, © 2010 Atlantis Press/World Scientific
- 22. 2 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon nical services to telecommunication services, business process modeling and popular web services. In the information system realm, this research effort has focused to a significant extent on (i) services definition: linguistic constructs and models to define and represent services’ behaviors and properties, (ii) services discovery: architectures or protocol suites to allow service sharing and functional matching and (iii) services composition: orchestra- tion of service components into more complex processes [2, 3, 4, 5]. Open computing environments created the needs for virtual cooperating systems to allow resource sharing. Digital enterprise ecosystems emerged as a concept for capturing the in- teractions of business networks. Ecosystems can comprise autonomous organizations and related services, sharing agreements on overall domain specific components and rules gov- erning interactions and inter-relationships among the participants. Enterprise ecosystems provide some formalization of common models, shared knowledge and global resources to enable loosely coupled interoperability among enterprises. In essence, they can be used, as opposed to open environments, to provide controlled business and enterprise environ- ments delimiting the collaboration scopes to a set of actors respecting business related rules. Ecosystems form a suitable environment for application composition. They provide an environment with identified semantics and business properties wherein peers providing services interact based on a global but not too restrictive agreement. This helps in distin- guishing functional needs, relations between them and other business-relevant properties. 1.1.1 Objectives and Contributions In this chapter we address the service-based application composition issue in a peer-to- peer ecosystem. In such an ecosystem, our approach consists of first defining a high level interaction between actors, then refining it to an application defined as a graph of abstract services. The application is realized by substituting abstract services for matching services provided by peers belonging to the ecosystem. We show how the concrete application real- ization based on service composition can take advantage of the ecosystem’s network reor- ganization into peer communities, in terms of communities’ definition and communication protocol by building on top of an unstructured system a hybrid overlay network. The remainder of the chapter is organized as follows. Section 1.2 exposes literature back- ground, namely service orientation and P2P systems and communities. Section 1.3 exposes ecosystems, peer-communities and services under a multi-layered comprehensive frame- work for service-based application composition. Section 1.4 focuses on the fourth layer of the framework, describing its organization in a super-peer based overlay network. Sec-
- 23. A Community-based Approach for Service-based Application Composition in an Ecosystem 3 tion 1.5 presents the European Electricity Exchange Market as an ecosystem example. It compares two views of the studied ecosystem, focusing on the process of application real- ization. Section 1.6 concludes the chapter and presents future work. 1.2 Background In this Section we discuss two recent developments that are changing the way IT appli- cations are designed, deployed and exchanged: (i) service oriented computing, providing a new paradigm for creating applications on demand and (ii) peer-to-peer systems, often used for sharing resources. We first describe current work in service oriented computing, then we briefly define P2P systems and review P2P communities-related literature. 1.2.1 Service Orientation Previous work in service oriented systems has focused to a significant extent on 1) con- structs and models to define and represent the behaviors and properties of services, and 2) the architectures or protocol suites to allow service sharing and matching and on ser- vices’ composition into more complex systems. A service can be viewed as a self-contained, modular basic software unit that is described, published and invoked over a network to create new software components or products. It encapsulates functions and modules of an application domain (e.g., business process com- ponents, supply chain units). It provides an interface to allow external invocation. Among service description models proposed in the literature, the Web Service Description Lan- guage (WSDL) [2] has become a de-facto industry standard. It is an XML-based model that allows a syntactical representation of the methods and parameters needed to inter- act with a service. Other models extend the syntactic representation of services by adding semantics to resolve definition discrepancies and heterogeneities that can hinder service matching and composition. For example, the METEOR-S project [3] extend WSDL with semantic annotations while the OWL-S [4] and WSMO [5] (the Web Service Modeling Ontology) approaches are based on an ontology of web services. The ontology provides a precise description of service components and their inter-relationships. Several standards and architectures are proposed to enable the integration and sharing of heterogeneous ser- vice. For example, Service Oriented Architecture (SOA) is a “paradigm for organizing and utilizing distributed capabilities that may be under the control of different ownership do- mains” [6, 7].
- 24. 4 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon Service discovery is defined by Keller et al. [8] as the automatic localization of services cor- responding to user’s need. Booth et al. [9] describe the discovery process as the localization of a machine readable description corresponding to given functional needs. Toma et al. [10] define service discovery as a process taking as input a user query and returning as output a list of available resources corresponding to the user’s need expressed in the input query. Two major aspects are tackled by service discovery, namely service localization and ser- vice matching. Service localization relies on either centralized or distributed architectural models. The UDDI (Universal Description Discovery and Integration [11, 12, 13]) became a widely know standard for centralized service localization. It consists of a set of UDDI nodes collaborating to create a global structure. Srinivasan et al. [14] extended the UDDI model to support OWL-S semantic description allowing more efficient comparison between the user’s need and the available services. Distributed localization models consisted first in setting up a distributed federation of UDDIs [15, 16]. Verma et al. [17], Paolucci et al. [18] and Schmidt et al. [19] discussed other complex models. Service matching is widely addressed in the literature. Ernst et al. [20] and Dong et al. [21] studied syntactic similarity based on trace data and clustering respectively. Paolucci et al. [22], Benatallah et al. [23] and the WSMO workgroup [8] tackled the matching based on semantic similarity. Taher et al. [24] and Bordeaux et al. [25] studied other approaches based on abstract services and labeled transition systems respectively. Service composition designates the interaction taking place between two or more services in order to accomplish a given goal. The composition process tackles several aspects, such as the interaction description and organization, the message exchange management, the transaction like behavior, the interaction context, the level of automation, the failure recov- ery, etc. The Web Services Business Process Execution Language (WS-BPEL [26]) is the current standard for describing services’ compositions. It allows to model compositions as interaction workflows. An alternative for BPEL is the Web Service Choreography Interface (WSCI [27]). Both BPEL and WSCI allow static service composition wherein services are bound at design time. Thakkar et al. [28], Casati et al. [29] and Sun et al. [30] present dynamic composition environments based on composition engines capable of binding se- lected services at runtime. The automation level of the composition process is also widely studied in the literature. An exhaustive survey on service composition is out of the scope of this work. Useful information is available in [31, 32].
- 25. A Community-based Approach for Service-based Application Composition in an Ecosystem 5 1.2.2 P2P Systems Peer-to-peer (P2P) systems are distributed systems composed of distinct computing ele- ments, called peers, with similar resources and capabilities. Peers interact together to share services and resources. P2P systems can be classified into unstructured and structured sys- tems. In unstructured P2P systems, peers are organized in random graphs with no control over their contents. Each peer controls its contents and the access and sharing of its re- sources. Unstructured P2P systems can be further classified into (i) centralized systems when a central directory is used to store global state information (indexes, data locations, etc.), (ii) decentralized systems when no global state information (network state informa- tion, context data) is maintained by the peers, and (iii) hybrid systems which combine the characteristics of centralized and decentralized by using super-nodes (or super-peers) [33] to control simple peers with less resources and capabilities. Structured P2P systems keep a tight control over network topology and peer contents by placing data not randomly in peers but at specific locations defined by the overlay network strategy (an indexing strategy). P2P systems can also be structured by using clustering techniques to group peers based on common properties or interests. Clusters can be viewed as communities belonging to overlays defined on top of unstructured P2P systems. According to Khambatti et al. [34], a community is a set of active peer members, involved in sharing, communicating and pro- moting common interests. Significant research is currently targeted at creating community- oriented overlay networks in order to avoid query messages flooding and to save resources in handling irrelevant queries over the P2P network. DHT-based techniques [35, 36] guar- antee location of content within a bounded number of hops by tightly controlling the data placement. Other techniques based on clustering strategies have been proposed to reduce query traffic, grouping peers sharing similar properties. According to Oztopra et al. [37], two main strategies are used the literature for clustering peers. The first strategy takes into account network related characteristics while the second focuses on peers interests. In the following we review both strategies considering that peers participating in a services’ ecosystem are mainly interested in providing, sharing and re- using services. 1.2.2.1 Using Network Characteristics to Build Peer Communities Several research was conducted on clustering of peers based on network characteristics. Ratnasamy et al. [38] present a scheme whereby nodes partition themselves into groups called bins such that nodes that fall within a given bin are relatively close to one another in
- 26. 6 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon terms of network latency. Zhang et al. [39] propose a topology aware system constructing an overlay network by exploiting the locality in the underlying network using the group concept. Each host in the overlay is running a protocol to communicate with other hosts. In general, each host maintains information about a set of other hosts to communicate with. Two hosts are considered as neighbors if they are connected through the overlay. MetaStream [40] is a content discovery protocol for topology-aware on-demand streaming. In MetaStream, clients choose streaming sources based on network distance. For this pur- pose, they self-organize into a dynamic hierarchy of clusters based on the network topology. Any protocol for constructing a topology-aware hierarchy can be used. Connectivity-based Distributed node Clustering (CDC) [41] implements node clustering based on node con- nectivity in P2P networks, while Zheng et al. [42] use an approach based on link delay of node communications in the P2P network. Oztopra et al. [37] propose to cluster peers based on time (communication duration) closeness. Disregarding the specific adopted technique, building communities based on network char- acteristics generates an overlay where peers in a community provide different services. In such a scenario, it is highly probable that peers in a community will behave in a co- operative manner. When one peer is selected, the possibility of selecting another member of the community is increased. This makes sense considering that it is better for a peer to search for a service among his neighbors before searching among further members. 1.2.2.2 Using Service Properties to Build Peer Communities Service properties are classified in two main categories: (i) functional and (ii) non- functional [5, 43]. Functional properties represent the functionality provided by the service and its semantic description elements, for example the related input/output parameter list (and conditions if available). Note that the service as a software unit might provide several functionalities. In this case, each functional aspect can be studied as a separate entity. Building communities based on the functionalities provided, allows us to obtain competitive communities, where each peer holds services accomplishing the same task, although some service attributes may vary. This way, each peer will compete with others to get selected by a client. The client choice is based on non-functional properties, which are not directly related to the functionality provided by the service. An exhaustive list or classification of those properties is out of the scope of our project, we note that some of the non-functional properties are QoS related and thus in correlation
- 27. A Community-based Approach for Service-based Application Composition in an Ecosystem 7 with the network characteristics such as execution time. Other non-functional properties do not express QoS but might form substrate criteria to build communities on, for instance security-level and trust [44, 45]. 1.3 A Framework for Sharing Services In the following we present and model peers interactions in a services’ ecosystem. We de- fine peer-communities in such ecosystems and formalize interaction rules. We also describe the multi-layered service-based composition framework under which on-demand applica- tion composition takes place. 1.3.1 Ecosystem, Peer-communities and Services In an ecosystem various communities organize business-driven collaboration among groups of service-providing peers. We first describe such ecosystems, and then we provide more precise definitions of the relevant terms. %HORQJVWR %HORQJVWR 3DUWLFLSDWHVLQ 3HHU 1L 0XVWDJUHHZLWKWKH HFRVVWHPJOREDO DJUHHPHQW *ORE$ ( 3HHURPPXQLW 3/RF$ 0XVWDJUHHZLWKWKHHFRVVWHP JOREDODJUHHPHQW*ORE$ ( 0XVWGHILQHDORFDODJUHHPHQW (FRVVWHP ( 0XVWGHILQHDJOREDO DJUHHPHQW *ORE$ ( 0XVWGHILQHDVHWRI J J /RF$ 3 DEVWUDFWVHUYLFHV$ 6HUYLFHV 3URYLGHV $EVWUDFW 6HUYLFHV $FFHVVHGYLD ,PSOHPHQW 'HILQHV 6HUYLFHV $EVWUDFW 6HUYLFHV S Figure 1.1 Ecosystem’s Organization As described in Figure 1.1, an ecosystem is a group of peer-communities in which each peer-community accepts a consensual specification of a business area and its related busi- ness rules, referred to as global agreement. The ecosystem defines a set of abstract services based on the global agreement. A peer respecting the global agreement provides services as implementations of the abstract ones. Peer-communities are groups of peers having a consensual agreement on a minimum set of properties. We denote by local agreement this set of the required properties. Each peer must satisfy its ecosystem’s global agreement and its communities’ local agreements.
- 28. 8 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon We model the ecosystem’s organization by the following definitions and rules: • Given an ecosystem E, its global agreement specification is the set of business-related properties denoted by GlobA(E). GlobA(E) = {pp}p=[1,...,q] where q is the number of relevant business properties in the ecosystem. For example response time, which is a widely used property. Properties notation language is chosen depending on the ecosystem domain(s). • Abstract services are defined by the ecosystem in order to disseminate an application domain knowledge. An abstract service is an interface defining the abstract operation needed to fulfill the related functionality. An abstract service describes the operation via a semantic business-related description, including but not limited to, its inputs, outputs and an associated set of constraints, typically restrictions. It provides no real implementation of the operation, just the signature. The abstract service interface al- lows to define also realization constraints to be respected by the interface implementer. An abstract service is designated by Ai. We denote by A the set of abstract services de- fined by the ecosystem E, and by n the number of described abstract services in the ecosystem such as |A| = n. • For each abstract service Ai the set of defined properties Pi is defined as: Pi = {pm i }m=[1,...,qi],i∈{1,...,n} subject to qi ⩽ q and ∃ f : pm i → pp. For instance, based on the business property response time, the abstract service’s interface defines the prop- erty execution time. We note that some business related properties might not be relevant for a given abstract service, thus they are not used in its interface. • Abstract services defined by an ecosystem E must comply with the GlobA(E) specifi- cation: ∀Ai ∈ A, Pi ⊆ GlobA(E) where i ∈ {1,...,n} • Services are defined by peers in order to present their business offers in the ecosys- tem. They are defined with respect to the ecosystem required functionalities, thus a service is a concrete implementation of an abstract one. A service Sij implementing an abstract service Ai must redefine the abstract operation and respect all the associ- ated constraints. Although having similar functional interfaces, two services Sij and Sij may differ in their non-functional properties. • For each service Sij the set of service-related properties is derived from the correspond- ing abstract service properties and is denoted by Pij = {pp ij}i∈{1,...,n}, p=[1,...,qi], j∈N. Actually a service redefines and implements the properties of its related abstract ser- vice. For example, the service redefines the property execution time inherited from its
- 29. A Community-based Approach for Service-based Application Composition in an Ecosystem 9 corresponding abstract service and evaluates it. If a service’s WSDL description pro- vides several operations related to different functionalities, the service is mapped to the required number of abstract services. • Within partnerships, services are offered as implementations of abstract services. An implementation relation IMPD is defined in order to associate a concrete service with its corresponding abstract service. The implementation dependency is such that: ∀Sij, ∀Ai ∈ A, i ∈ {1,...,n}, j ∈ N IMPD(Sij,Ai) =⇒ ∀ pp i ∈ Pi, ∃ pp ij ∈ Pij where p, p ∈ {1,...,qi} • Given a peer-community PC, its local agreement specification is denoted by LocA(PC). A local agreement is specified either in terms of services properties or any other criteria relevant to the studied ecosystem members (e.g. locality, peer-trust-level, etc.). • Peers belonging to a community PCLocA must comply with the local agreement speci- fication LocA(PC). ∀Ni ∈ PCLocA, Ni respects LocA(PC) For instance, given a community PCLocA based on the local agreement LocA : equal trust-level, all its member peers share the same value for the property trust-level. • We denote by S(Ni) the set of services provided by the peer Ni and by S(PC) the services available in the community PC. S(PC) is the union of the services whose providing peers comply with the local agreement LocA(PC). 1.3.2 Multi-layered Service-based Composition Framework The multi-layered framework for service-based application composition is illustrated in figure 1.2. It allows dynamic application composition in a given ecosystem. It is composed of five layers. The first layer models the studied business logic in a workflow of activities from which abstract services are described. The second layer allows to define an applica- tion modeled by a graph of abstract services. The third layer contains the set of realizations of the application. A realization is defined as a combination of available services on the net- work capable of executing the application process. The fourth layer is the virtual overlay network in which peers are clustered in communities 1. The fifth layer represents the under- lying peer architecture. At this layer we capture peer related non-functional characteristics that help assessing network related measures. The service binding is deferred until runtime, 1For simplification purposes, we do not distinguish hereafter between the terms peer-community and community.
- 30. 10 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon ďƐƚƌĂĐƚ^ĞƌǀŝĐĞƐ KŶŽƚůŽŐŝĞƐͬ ĞƐĐƌŝƉƚŝŽŶƐ ƵƐŝŶĞƐƐĐƚŝǀŝƚŝĞƐ %XVLQHVV /DHU ϭ Ϯ Ŷ ͙ Đϭ Đϯ ĐϮ ĐŶ KŶŽƚůŽŐŝĞƐͬĞƐĐƌŝƉƚŝŽŶƐ 'W $SSOLFDWLRQ /DHU ^ĞƌǀŝĐĞ ŽŵƉŽƐŝƚŝŽŶ 6HUYLFH 5HSRVLWRU ŽŽŬŬĞĞƉŝŶŐ ͗ /ϭ /Ϯ /ŵ ͙ ,QVWDQFHV /DHU ŽŵŵƵŶŝƚLJ ŝƐƚŝŶŐŽĨ ĐŽŶĐƌĞƚĞƐĞƌǀŝĐĞƐ ĨŽƌĞĂĐŚ/ŶƐƚĂŶĐĞ ŽŵŵƵŶŝƚLJ ŽŵŵƵŶŝƚLJ ŽŵŵƵŶŝƚLJ 2YHUOD 1HWZRUN /DHU ŝƌĞĐƚŽƌLJ ŽĐĂůƐƚĂƚĞ ƚĂďůĞƐ 8QGHUOLQJ 1HWZRUN /DHU EŽĚĞƐ/ŶƚĞƌĐŽŶŶĞĐƚŝǀŝƚLJĂƌĐŚŝƚĞĐƚƵƌĞ ;hŶƐƚƌƵĐƚƵƌĞĚƵŶĚĞƌůLJŝŶŐWϮWĂƌĐŚŝƚĞĐƚƵƌĞͿ WƌŽǀŝĚĞƌWĞĞƌƐůŝƐƚŝŶŐƐ Figure 1.2 Multi-layered Composition Framework allowing a dynamic cost-based service selection. In the following we briefly describe the framework layers and components. • The Business Layer models the business logic in a workflow of required activities. The main purpose is to refine business activities in abstract services modeling func- tionalities shared by applications in the domain of interest. • The Application Layer represents the composite application by a graph of abstract services denoted by Generic Business Process (GBP). A GBP is an oriented attributed graph whose vertices represent abstract services and edges represent control sequences indicating functional dependencies between the abstract services. Attributes are associ- ated with the vertices and the edges in order to represent functional and non-functional data and characteristics. Yetongnon et al. [46] discusses details about this layer and the following ones. • The Instances Layer contains a set of possible service compositions generated from the GBP abstract service graph and based on the available services in the ecosystem. This conversion of a GBP into a set of GBP instances is carried out by an instantia- tion process in which services registered by peers are substituted for the abstract ser- vices of the GBP. Thus, a GBP instance is a directed attributed graph whose nodes are registered services, edges connect two services based on the functional dependencies
- 31. A Community-based Approach for Service-based Application Composition in an Ecosystem 11 expressed in the GBP, and the attributes values are derived from the corresponding attributes of both nodes and edges in the GBP graph. The study of the instantiation process is out of the scope of this chapter. • The Overlay Network Layer is the peers’ organization into a community-based over- lay network. The overlay description, organization and communication is detailed the following section. • The Underlying Network Layer helps capturing the underlying network characteris- tics. At this point, the services properties can be evaluated along with the properties of the edges connecting the hosting peers in an instance graph. Peer properties are projected on the corresponding instances graphs. Each enterprise is modeled by a set of peers such as each enterprise application server, providing services or requiring an application instantiation, is a peer. • The Service Repository component interacts with the five layers. It provides at each layer the required elements (cf. figure 1.2). For instance at the business layer, it con- tains the ontologies and the abstract services listings. 1.4 The Overlay Network The overlay network is a view of the ecosystem filtered by peer-communities local agree- ments. For instance, local agreements consisting of the property providing similar function- alities generate an overlay of peer-communities in which peers providing services imple- menting the same abstract service are regrouped in the same community. 1.4.1 Overlay Organization We adopt the classical two levels peer organization,consisting of peer groups each managed by a super-peer. Figure 1.3 illustrates an example of a super-peer based overlay network architecture for service oriented application development. It consists of peers whose main goal is to provide concrete implementations of abstract services. Peers are organized in communities managed by super-peers which are in turn organized in a communication topology. For example, peer-community 1 is managed by super-peer SN1 and includes four peers N1,...,N4. Note that a peer can provide implementations for one, several abstract services or none; on the other hand an abstract service can be implemented by more than one peer. A peer-community is a set of peers respecting a local agreement and managed by a super-
- 32. 12 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon ďƐƚƌĂĐƚƐĞƌǀŝĐĞϭ ďƐƚƌĂĐƚƐĞƌǀŝĐĞϮ Eϭ ^EϮ WĞĞƌͲŽŵŵƵŶŝƚLJϮ WĞĞƌͲŽŵŵƵŶŝƚLJϭ ďƐƚƌĂĐƚƐĞƌǀŝĐĞϯ ďƐƚƌĂĐƚƐĞƌǀŝĐĞϰ EϮ Eϯ ^Eϭ ^Eϯ WĞĞƌ ŽŵŵƵŶŝƚLJ ϯ ďƐƚƌĂĐƚƐĞƌǀŝĐĞϱ Eϰ ^Eϯ /ŵƉůĞŵĞŶƚƐ ŽŐŝĐĂů ůŝŶŬ WĞĞƌͲŽŵŵƵŶŝƚLJϯ ĐŽƐLJƐƚĞŵ /ŵƉůĞŵĞŶƚƐ ŽŐŝĐĂůůŝŶŬ ĐŽƐLJƐƚĞŵ Figure 1.3 Example of the Overlay Network Organization peer. Formally, it is denoted by PCLocA = SNk, {Nr}r=1,...,nk where SNk is the super-peer managing the set of nk peers respecting the local agreement LocA(PC). Peer-communities are based on either (i) network characteristics or (ii) proper- ties connected to services as described in section 1.2.2. 1.4.2 Super-peers Super-peers are selected based on their computing capabilities (in order to handle the GBP instantiation) and/or their trustworthiness2. The links between super-peers are chosen as the shortest path from the physical network. Each link between super-peers represents a bidirectional communication path. Super-peers in the overlay network maintain and manage a distributed directory structure. Each super-peer maintains a local repository, consisting of two tables: a local state infor- mation table (for example tables 1.3(a), 1.3(b), 1.4(a)) and a global state information table (for example tables 1.2, 1.4(b)). The local state information table contains: (i) the set of peers managed by the super-peer. (ii) A state St(Nr) for each peer. St(Nr) =ON if the peer Nr is on-line. St(Nr) =OFF if the peer Nr is off-line. (iii) For each peer Nr, a list of provided services Sik and their related abstract services Ai. The global state information table, community directory, represents for each super-peer SNk in the overlay, the set of abstract services {Ai}i=1,...,ni that are supported by its community. 2For the sake of conciseness we will not detail the super-peers choice and we consider for the rest of the chapter that a super-peer does not depart.
- 33. A Community-based Approach for Service-based Application Composition in an Ecosystem 13 1.4.3 Event Related Communication Two major events need to be considered, first peer join and second peer departure. When a peer joins the ecosystem three actors or group of actors are implicated. First the peer itself, (i) launches a probing process to discover the closest super-peer in terms of physical distance then (ii) it queries the selected super-peer asking for the list of abstract services and for the community directory table. Then the peer (iii) decides on abstract services to implement 3 or if already implemented it grants network members access to its services. If needed, the peer creates mappings between its existing services and one or more related abstract services. The peer respects indirectly the ecosystem’s global agree- ment GlobA by choosing to implement an abstract service or by providing required map- pings for its existing services. Finally the peer (iv) sends requests to the super-peers of the communities it is willing to join, notifying them of its presence in the community. Clearly respecting the local agreement LocA(PC) of each of the solicited communities is a join pre- requisite. Second, each of the concerned super-peers (i) receives the joining peer request and information, (ii) updates its local copy of the community directory and (iii) sends up- date notifications to direct super-peers neighbors. Third, other super-peers (i) receive the community directory update notifications and (ii) proceed on updating their global state information. When a peer departs, the same actors are implicated. First, the peer itself notifies its super- peers before going offline. We adopt clean peer departure considering that peers main goal is to collaborate, improving the network and its added value (generated applications). Sec- ond, each of its related super-peers (i) flags the peer as offline in the local state table. After- ward, if the departing peer is the last to provide a given functionality, (ii) the related abstract service is removed for the community directory. Finally the super-peer (iii) sends update notifications to neighbors super-peers containing the community directory new state. Third, other super-peers (i) receive the community directory update notifications and (ii) proceed on updating their global state information. 1.5 Case Study: The European Electricity Market Produced electricity cannot be stored for long, therefore the market must undergo a regula- tion process. Market regulation consists of insuring that the quantity of produced electricity is equal to the needed consumption power. Electricity regulation is ensured via exchanges 3respecting the corresponding implementation relation IMPD.
- 34. 14 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon between national transmission system operators (TSOs) in Europe. Exchanges are under control of the UCTE (Union for the Co-ordination of Transmission of Electricity) 4. TSOs are members of UCTE, i.e., RTE in France, RWE-TSO in Germany and REE in Spain. A TSO is also in charge of electricity transportation to large consumers (for example national railway systems) and to suppliers, including the high voltage to medium voltage transfor- mation required for suppliers. A supplier (e.g. Poweo, DirectEnergie in France) delivers electricity to individual consumers. A supplier receives medium voltage electricity from a TSO and insures its transformation to low consumer-compliant voltage. An individual consumer is the system’s end-user, it receives low voltage electricity from a supplier after a MV-LV transformation. Suppliers decide which and how many consumers are cut-off in case needed. A TSO is also in relation with one or several producers. National electricity industries (e.g. EDF in France, Alsthom in Germany, Landsvirjun in Island) produce and sell electricity to their national TSO. A producer is in relation with one or several TSO’s. Industrial companies and electricity suppliers are consumers of the TSOs. These actors contractually bound form our ecosystem. (XURSHDQ/DHU 1DWLRQDO/DHU )UDQFH 1DWLRQDO/DHU *HUPDQ 3URGXFWLRQ ([FKDQJH PDQDJHPHQW 6XSSO PDQDJHPHQW RQVXPSWLRQ EXVLQHVVD[LV $FWLYLWLHV D[LV 87( (') :LQG 57( 3RZHR(') 'LUHFW(QHUJLH /DUJHFRQVXPHUV HJ61) $/67+20 5:(762 9DWWHQIDOO /DUJHFRQVXPHUV ,QGLYLGXDO FRQVXPHUV ,QGLYLGXDO FRQVXPHUV ([FKDQJHXQGHU FRQWUDFW IRU EDODQFLQJ (XURSHDQ/DHU 1DWLRQDO/DHU )UDQFH 1DWLRQDO/DHU *HUPDQ 3URGXFWLRQ ([FKDQJH PDQDJHPHQW 6XSSO PDQDJHPHQW RQVXPSWLRQ EXVLQHVVD[LV $FWLYLWLHV D[LV 87( (') :LQG 57( 3RZHR(') 'LUHFW(QHUJLH /DUJHFRQVXPHUV HJ61) $/67+20 5:(762 9DWWHQIDOO /DUJHFRQVXPHUV ,QGLYLGXDO FRQVXPHUV ,QGLYLGXDO FRQVXPHUV ([FKDQJHXQGHU FRQWUDFW IRU EDODQFLQJ 3URGXFWLRQ ([FKDQJH PDQDJHPHQW 6XSSO PDQDJHPHQW RQVXPSWLRQ EXVLQHVVD[LV $FWLYLWLHV D[LV 87( (') :LQG 57( 3RZHR(') 'LUHFW(QHUJLH /DUJHFRQVXPHUV HJ61) $/67+20 5:(762 9DWWHQIDOO /DUJHFRQVXPHUV ,QGLYLGXDO FRQVXPHUV ,QGLYLGXDO FRQVXPHUV ([FKDQJHXQGHU FRQWUDFW IRU EDODQFLQJ +LJKYROWDJHSRZHU VXSSO /RZ YROWDJHHOHFWULFLW VXSSO (OHFWULFLW VDOHV 0HGLXP YROWDJHSRZHUVXSSO +LJKYROWDJHSRZHU VXSSO /RZ YROWDJHHOHFWULFLW VXSSO (OHFWULFLW VDOHV 0HGLXP YROWDJHSRZHUVXSSO Figure 1.4 Activity flows and Ecosystem actors The main activities of the ecosystem are the electricity production, the exchange manage- ment and the management of electricity supply and consumption. We have implemented our ecosystem along two axes: activities axis and business axis (figure 1.4). This figure illustrates : 4www.ucte.org
- 35. A Community-based Approach for Service-based Application Composition in an Ecosystem 15 - on its horizontal axis, the business process in terms of sequence of activities; - on its vertical axis, an example of the enterprises involved in the functionalities of the ecosystem; - an example of interaction between the described actors is also represented and - finally these enterprises are also divided geographically. Communities appear based on the activities axis elements but also in terms of the business axis components. It is clear that location is important in this ecosystem because electric- ity transportation depends on the physical network. For example, Spain can call-in only Portugal and France for its electricity regulation. To illustrate our approach, consider a simplified example of the European Electricity Mar- ket activity as described in figure 1.4. From the activities production and supply manage- ment we define respectively the production and the delivery abstract services. The activity exchange management is refined in two abstract services: regulation and transformation. The studied application models the electricity regulation via production and supply man- agement. In this process the consumer is a passive actor, therefore we do not model the con- sumption activity in the studied application. The application GBP and the services along with their providing peers are illustrated in Figure 1.5. „•–”ƒ –‡”˜‹ ‡• ‡”˜‹ ‡•ƒ†”‘˜‹†‡”‡‡”• ’”‘†— –‹‘ ”‡‰—Žƒ–‹‘ –”ƒ•ˆ‘”ƒ–‹‘ †‡Ž‹˜‡”› Ƈſ Ş’”‘†ř ƀřſ Ş’”‘†řƀƈ Ƈſřƀřſř ƀřſřƀřſř ƀƈ Ƈſ‘™‡‘řƀřſ‹”‡ –‡”‰‹‡řƀřſƒ––‡ˆƒŽŽř ƀƈ Ƈſ ކ‡Žř ƀřſ ކ‡Žřƀƈ 3URGXFWLRQ 5HJXODWLRQ 7UDQVIRUPDWLRQ 'HOLYHU Figure 1.5 Example of a GBP, abstract services and provider peers Using this scenario, we present hereafter two examples of community construction based on, respectively, regional locality and functionalities provided. 1.5.1 A Regional Locality-based Overlay Figure 1.6 represents an overlay network based on regional locality, following the ver- tical axis of the ecosystem representation in figure 1.4. It consists of five super-peers and their related peer-communities. Each super-peer manages a community composed of peers, coming from the same geographical region. For instance the community managed by SN2
- 36. 16 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon Table 1.1 Concrete Services and related Execution Prices Service Price($) Service Price($) EDF-prod 0.1 Poweo 0.5 ALSTHOM-prod 0.2 DirectEnergie 0.4 RTE 0.15 Vattenfall 0.45 RWETSO 0.2 EDF-del 0.8 REE 0.3 ALSTHOM-del 1.2 UTCE 0.25 contains peers belonging to Germany, whereas the community managed by SN4 contains peers from the French-German border. Peers can belong to several communities, for exam- ple: peer B belongs to the community managed by SN2 and to the one managed by SN4. The links represent a bidirectional communication, whose cost is a communication unit if the exchange occurs between two super-peers and half a unit if a super-peer is communi- cating with one peer in its community. For example an invoke-reply between SN3 and H costs 0.5; a query-answer (or send-acknowledge) between SN3 and SN1 costs 1. ^Eϰ ^Eϯ ^EϮ ^Eϭ ^Eϱ , ͲĚĞů ͲƉƌŽĚ WŽǁĞŽ ^,dKDͲĚĞů ^d,KDͲƉƌŽĚ ŝƌĞĐƚŶĞƌŐŝĞ ' sĂƚƚĞŶĂůů hd , ͲĚĞů ͲƉƌŽĚ Z Ztd^K ŝƌĞĐƚŶĞƌŐŝĞ ^,dKDͲĚĞů ^d,KDͲƉƌŽĚ Zd WŽǁĞŽ Figure 1.6 Locality-based view of the studied ecosystem The community directory shown in table 1.2 is based on the overlay illustrated in figure 1.6. This directory is updated on a regular basis. For example after peer C departure the commu- nity managed by SN4 no longer provides an implementation for the functionality transfor- mation, thus transformation is removed from the list of supported abstract services by the community of SN4 in the community directory shown in table1.2, whereas no changes are
- 37. A Community-based Approach for Service-based Application Composition in an Ecosystem 17 made to the community directory entry related to SN5, because the functionality transfor- mation is still implemented by peer E. Both SN4 and SN5 flag H as OFF in their respective local state tables 1.3(a) and 1.3(b). SN4 and SN5 notify the remaining super-peers sending them incremental updates of the community directory. Table 1.2 Community Directory in the overlay based on regional locality Super-peer Supported Abstract Services SN1 {production, regulation, transformation, delivery} SN2 {production, regulation, transformation, delivery} SN3 {production, regulation, delivery} SN4 {production, transformation, delivery} SN5 {transformation} Consider the following user requirements where the GBP in Figure 1.5 is instantiated by super-peer SN4: (i) give preference to services in the community of SN4 (ii) minimize the total execution price in terms of execution cost in $ as shown in Table 1.1. From the local state table (Table 1.3(a)) SN4 determines the abstract services implemented by local mem- bers, in this case we have respectively for production: (ALSTHOM-prod, B); for transfor- mation: (DirectEnergie, C) and for delivery: (ALSTHOM-del, B). Functionalities production, transformation and delivery are matched locally while the regulation functionality must be matched remotely by a neighbor super-peer. On SN4 the GBP is expressed as the union of a local and a remote subsets of abstract ser- vices. The local subset contains abstract services implemented by the community members, whereas the remote subset designates the set of abstract services whose implementations will be discovered on the network. To compute the local part of the instance, SN4 asks its underlying peers for the execution price (table 1.1) of the services matching the abstract services in the local part of the GBP. In the chosen example each functionality is matched by one concrete service, thus the following services are selected to compose the starting combination: {(ALSTHOM-prod, B), ((DirectEnergie, C), (ALSTHOM-del, B)}. To compute the remote part of the instance SN4 uses table 1.2 to find other super-peers whose commu- nities implement the regulation abstract service. SN4 queries SN1, SN2 and SN3. SN1 and SN3 reply respectively with (RTE, A, 0.15), (REE, D, 0.3); while SN2 replies with both (UTCE, G, 0.25) and (RWETSO, F, 0.2). Aiming on minimizing the execution cost in $, SN4 selects (RTE, A, 0.15). On SN4, the GBP starting instance becomes {(ALSTHOM-prod, B), (RTE, A), (DirectEnergie, C), (ALSTHOM-del, B)}. This process is available in [46].
- 38. 18 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon Table 1.3 Example of Local state tables on super-peers (overlay based on regional locality) (a) Local State Table on SN4 Peer (Service, Functionality) PID St (Ni) B ON {(ALSTHOM-prod, production), (ALSTHOM-del, delivery)} C ON {(DirectEnergie, transformation)} (b) Local State Table on SN5 Peer (Service, Functionality) PID St (Ni) C ON {(DirectEnergie, transformation)} E ON {(Poweo, transformation)} In the following we compute the communication cost of the instantiation process on SN4. The intra-community communication cost is due to query-response communications be- tween SN4 and the peers B and C. SN4 requests a service execution time from each of its peers, generating a communication cost of 1. The inter-communities communications are generated by communications between SN4 and the super-peers SN1, SN2 and SN3. The total communication cost between super-peers is equal to 3. Note that SN4 did not query SN5 taking into consideration that the community of SN5 does not provide imple- mentations for regulation (Table 1.2). Each of SN1 and SN3 queried one member peer, while SN2 queried two member peers, thus the total inter-community cost is equal to 3 + 0.5 + 0.5 + (2 ∗ 0.5) = 5. Finally the service invocation cost corresponds to the sum of the invocation costs of each of the services belonging to the starting instance. In this example we consider the invocation cost of a service equal to the cost of the communica- tion with its hosting peer, hence the total service invocation cost for the starting instance equals 3. In fact SN4 contacted C once (0.5), B twice (1) and A once via SN1 (1.5). Note that the starting instance predicted cost in dollar is equal to 1.95$. 1.5.2 A Functionality-based Overlay Figure 1.7 illustrates an overlay network based on functionalities. Each peer-community implements an abstract service: member peers provide implementations of the abstract ser- vice interface with different non-functional properties. The links represent the two-way communication between peers as described in section 1.5.1. Four peer-communities are built based on the functionalities shown in Figure 1.5. For example, the super-peer SN1 manages the community related to the abstract service production; member peers, B and H provide an implementation to production. Note that a member peer might provide im-
- 39. A Community-based Approach for Service-based Application Composition in an Ecosystem 19 plementations for other abstract services, thus it will belong to several peer-communities. For instance, peer B provides implementations for both production and delivery, therefore it belongs to two peer-communities: one managed by SN1 and related to the functionality production and the other managed by SN4 and related to the functionality delivery. For example, the communities managed by SN2 and SN3 are respectively denoted by PC(regulation) = (SN2, {A, D, F, G}) and PC(transformation) = (SN3, {C, E, G}). , ͲƉƌŽĚ ^d,KDͲ ƉƌŽĚ ŝƌĞĐƚŶĞƌŐŝĞ WŽǁĞŽ ' sĂƚƚĞŶĨĂůů , ͲĚĞů ^d,KDͲ ĚĞů ' hd Z Zd Ztd^K ^Eϭ ^Eϯ ^EϮ ^Eϰ Figure 1.7 Functionality-based view of the studied ecosystem Each super-peer manages a copy of the community directory shown in Table 1.4(b). In this example, Table 1.4(b) assists a super-peer when looking for an implementation of a required abstract service. When peer H goes off-line, SN4 updates its local state information table (Table 1.4(a)) by flagging H as OFF and removing (EDF-del, delivery). At the same time SN1 updates its local state information table by flagging H as OFF and removing (EDF-prod, production). Table 1.4 Service Repository tables in the overlay based on functionalities (a) Local state table on the super-peer SN4 Peer (Service, Functionality) PID St (Ni) B ON {(ALSTHOM-del, delivery)} H ON {(EDF-del, delivery)} (b) Community Directory on each super-peer Super-peer Supported Abstract Services SN1 {production} SN2 {regulation} SN3 {transformation} SN4 {delivery}
- 40. 20 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon Considering the same GBP on SN4, the user still wants to minimize the total execution price but he cannot choose to privilege services in his community anymore. Using Table 1.4(a), SN4 asks his member peers for the execution price of their provided services. SN4 selects locally (EDF-del, H) for the delivery functionality. Relying on Table 1.4(b), SN4 queries the other super-peers asking them for their best candidates in term of execution cost. SN4 receives from SN1: (EDF-prod, H), (RTE, A) from SN2 and (DirectEnergie, C) from SN3. On SN4, the starting GBP instance is {(EDF-prod, H), (RTE, A), (DirectEnergie, C), (EDF-del, H)}. Note that this starting combination costs 1.45$. Upon receiving SN4’s request, each of SN1, SN2 and SN3 engage local intra-community communication in order to select the best candidate. From the perspective of SN3 the total inter-communities communication cost is equal to 3+1+2+1.5 = 7.5. In fact SN4 queried each of SN1, SN2 and SN3 generating a communication cost equal to 3, SN1 generated 1 as communication cost by querying B and H for the execution prices; similarly SN2 and SN3 generated communication cost is respectively 2 and 1.5. As for the total intra-community local communication, SN4 queried two member peers generating a communication cost equal to 1. Finally the starting combination invocation cost is 5, SN4 contacted 3 super- peers (3) and one local member peer (0.5). Each of the contacted super-peer invoked one peer (3 ∗ 0.5). 1.5.3 Discussion Compared to the overlay based on regional locality, the organization based on functional- ities generates more traffic during the described phases of the instantiation process. This difference is due to the inter-communities message exchange which is higher because each abstract service is matched to a concrete one in a different community. While in the over- lay based on regional locality, the local community (instantiating the GBP) might provide services matching more than one functionality. Nevertheless the computed starting combi- nation in the overlay based on functionalities has a lower cost in $ compared to the one in the overlay based on distance. In fact, in this scenario, when the overlay is based on regional locality, the local cheapest services are selected whereas when based on functionalities, the cheapest service from each community is chosen. Those examples helped us to develop a phase of the instantiation process and to study the communication protocol. In reality, the constraints are more complex than selecting the cheapest service and the communication costs are not constant (they are function of bandwidth, quantity of data exchanged, etc.).
- 41. Bibliography 21 1.6 Conclusions and Further Research This chapter addressed service-based application composition in ecosystems. Given an ecosystem and a set of associated business-related properties, abstract services are de- fined to model business-relevant functionalities in the ecosystem. Peers complying with the ecosystem agreement provide services implementing the abstract services interfaces and respecting their related properties. We presented a multi-layered comprehensive framework for service-based application composition. The associated application composition approach, first consists of modeling the studied ecosystem’s business logic in a workflow of activities from which abstract ser- vices are described. Second, an application is defined as a graph of abstract services. Third, the set of application realizations is defined as a set of combinations of available services in the ecosystem capable of executing the application process. In order to improve peers communication effectiveness, we proposed to organize them into communities in a hybrid overlay network based on selected properties relevant to the ecosystem. This network organization helps us to abridge the discovery process and to focus more on the instantiation and the failure recovery activities. We described commu- nication related to the instantiation process in a case study of the European Electricity Ecosystem. We presented a comparison of a simplified application instantiation example in two different views of the ecosystem, namely, the regional locality view and the pro- vided functionalities view. The failure recovery process and the use of other non-functional properties as trust to build peer-communities will be addressed in future work. Bibliography [1] I. Vessey and A. P. Sravanapudi, Case tools as collaborative support technologies, Commun. ACM. 38(1), 83–95, (1995). ISSN 0001-0782. doi: http://doi.acm.org/10.1145/204865. 204882. [2] World Wide Web Consortium (W3C). Web Services Description Language (WSDL) Version 2.0. http://www.w3.org/TR/wsdl20/. [3] A. A. Patil, S. A. Oundhakar, A. P. Sheth, and K. Verma. Meteor-s web service annotation framework. In WWW ’04: Proceedings of the 13th international conference on World Wide Web, pp. 553–562, New York, NY, USA, (2004). ACM. ISBN 1-58113-844-X. doi: http://doi.acm.org/10.1145/988672.988747. [4] D. Martin, M. Burstein, D. Mcdermott, S. Mcilraith, M. Paolucci, K. Sycara, D. L. Mcguin- ness, E. Sirin, and N. Srinivasan, Bringing semantics to web services with owl-s, World Wide Web. 10(3), 243–277, (2007). ISSN 1386-145X. doi: http://dx.doi.org/10.1007/ s11280-007-0033-x. [5] D. Roman, U. Keller, H. Lausen, J. de Bruijn, R. Lara, M. Stollberg, A. Polleres, C. Feier,
- 42. 22 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon C. Bussler, and D. Fensel, Web service modeling ontology, Applied Ontology. 1(1), 77– 106, (2005). [6] OASIS. Reference model for service oriented architecture 1.0 (October, 2006). http://docs.oasis-open.org/soa-rm/v1.0/soa-rm.pdf. [7] M. J. Carey, SOA what?, IEEE Computer. 41(3), 92–94, (2008). [8] U. Keller, R. Lara, A. Polleres, I. Toma, M. Kifer, and D. Fensel. Wsmo web service discovery (November, 2004). URL http://www.wsmo.org/2004/d5/d5.1/v0.1/20041112/. [9] D. Booth, H. Haas, F. McCabe, E. Newcomer, M. Champion, C. Ferris, and D. Or- chard. Web services architecture, (2004). URL http://www.w3.org/TR/2004/ NOTE-ws-arch-20040211/. [10] I. Toma, D. Fensel, M. Moran, K. Iqbal, T. Strang, and D. Roman. An Evaluation of Discovery approaches in Grid and Web services Environments. In The 2nd International Confer- ence on Grid Service Engineering and Management, Erfurt, Germany (September, 2005). URL http://www.ict.swin.edu.au/conferences/gsem2005/. [11] T. Bellwood, S. Capell, L. Clement, J. Colgrave, M. J. D. D. Feygin, A. Hately, R. Kochman, P. Macias, M. Novotny, M. Paolucci, C. von Riegen, T. Rogers, K. Sycara, P. Wenzel, and Z. Wu. Uddi version 3.0.2 (October, 2004). URL http://uddi.org/pubs/uddi-v3. 0.2-20041019.htm. [12] T. Bellwood, L. Clment, D. Ehnebuske, A. H. M. Hondo, Y. L. Husband, K. Januszewski, S. Lee, B. McKee, J. Munter, and C. von Riegen. Uddi version 3.0 (July, 2002). URL http: //www.uddi.org/pubs/uddi-v3.00-published-20020719.htm. [13] T. Bellwood, D. Bryan, V. Draluk, D. Ehnebuske, T. Glover, A. Hately, Y. L. Husband, A. Karp, K. Kibakura, C. Kurt, J. Lancelle, S. Lee, S. MacRoibeaird, A. T. Manes, B. McKee, J. Munter, T. Nordan, C. Reeves, D. Rogers, C. Tomlinson, C. Tosun, C. von Riegen, and P. Yendluri. Uddi version 2.04 (July, 2002). URL http://uddi.org/pubs/ ProgrammersAPI-V2.04-Published-20020719.htm. [14] N. Srinivasan, M. Paolucci, and K. Sycara. Adding owl-s to uddi, implementation and through- put service. In First International Workshop on Semantic Web Services and Web Process Composition (SWSWPC 2004) 6-9, 2004, San Diego, California, USA, (2004). [15] P. Rompothong and T. Senivongse. A query federation of uddi registries. In ISICT ’03: Proceed- ings of the 1st international symposium on Information and communication technologies, pp. 561–566. Trinity College Dublin, (2003). [16] K. Sivashanmugam, K. Verma, and A. Sheth. Discovery of web services in a federated registry environment. In ICWS ’04: Proceedings of the IEEE International Conference on Web Services, p. 270, Washington, DC, USA, (2004). IEEE Computer Society. ISBN 0-7695- 2167-3. [17] K. Verma, K. Sivashanmugam, A. Sheth, A. Patil, S. Oundhakar, and J. Miller, Meteor-s wsdi: A scalable p2p infrastructure of registries for semantic publication and discovery of web services, Information Technology and Management. 6(1), 17–39 (January, 2005). URL http://dx.doi.org/10.1007/s10799-004-7773-4. [18] M. Paolucci, K. Sycara, T. Nishimura, and N. Srinivasan. Using daml-s for p2p discovery. In International Conference on Web Services (ICWS03), (2003). URL http://www.cs. cmu.edu/~{}softagents/papers/p2p_icws.pdf. [19] C. Schmidt and M. Parashar, A peer-to-peer approach to web service discovery, World Wide Web. 7(2), 211–229, (2004). ISSN 1386-145X. doi: http://dx.doi.org/10.1023/B:WWWJ. 0000017210.55153.3d. [20] M. D. Ernst, R. Lencevicius, and J. H. Perkins. Detection of web service substitutability and composability. In WS-MaTe 2006: International Workshop on Web Services — Modeling and Testing, pp. 123–135, Palermo, Italy (June 9,, 2006). [21] X. Dong, A. Halevy, J. Madhavan, E. Nemes, and J. Zhang. Similarity search for web services.
- 43. Bibliography 23 In VLDB ’04: Proceedings of the Thirtieth international conference on Very large data bases, pp. 372–383. VLDB Endowment, (2004). ISBN 0-12-088469-0. [22] M. Paolucci, T. Kawamura, T. R. Payne, and K. P. Sycara. Semantic matching of web services capabilities. In ISWC ’02: Proceedings of the First International Semantic Web Confer- ence on The Semantic Web, pp. 333–347, London, UK, (2002). Springer-Verlag. ISBN 3-540-43760-6. [23] B. Benatallah, M.-S. Hacid, A. Leger, C. Rey, and F. Toumani, On automating web services discovery, The VLDB Journal. 14(1), 84–96, (2005). ISSN 1066-8888. doi: http://dx. doi.org/10.1007/s00778-003-0117-x. [24] Y. Taher, D. Benslimane, M.-C. Fauvet, and Z. Maamar. Towards an Approach for Web services Substitution. In IDEAS ’06: Proceedings of the 10th International Database Engineering and Applications Symposium, pp. 166–173. IEEE Computer Society (Dec., 2006). ISBN 0-7695-2577-6. doi: http://dx.doi.org/10.1109/IDEAS.2006.45. URL http://liris. cnrs.fr/publis/?id=2464. [25] L. Bordeaux, G. Salaün, D. Berardi, and M. Mecella. When are two web services compatible? In eds. M.-C. Shan, U. Dayal, and M. Hsu, TES, vol. 3324, Lec- ture Notes in Computer Science, pp. 15–28. Springer, (2004). ISBN 3-540- 25049-2. URL http://springerlink.metapress.com/openurl.asp?genre= articleamp;issn=0302-9743amp;volume=3324amp;spage=15. [26] OASIS. Web Services Business Process Execution Language (WS-BPEL) Version 2.0. www. oasis-open.org/committees/wsbpel/. [27] World Wide Web Consortium (W3C). Web Service Choreography Interface. http://www.w3. org/TR/wsci/. [28] S. Thakkar, C. A. Knoblock, and J. L. Ambite. A view integration approach to dynamic com- position of web services. In In Proceedings of 2003 ICAPS Workshop on Planning for Web Services, (2003). [29] F. Casati, S. Ilnicki, L.-j. Jin, V. Krishnamoorthy, and M.-C. Shan. Adaptive and dynamic service composition in eflow. In CAiSE ’00: Proceedings of the 12th International Con- ference on Advanced Information Systems Engineering, pp. 13–31, London, UK, (2000). Springer-Verlag. ISBN 3-540-67630-9. [30] H. Sun, X. Wang, B. Zhou1, and P. Zou. Research and implementation of dynamic web services composition. In Advanced Parallel Processing Technologies. pringer Berlin / Heidelberg, (2003). [31] J. Rao and X. Su, A survey of automated web service composition methods, Semantic Web Services and Web Process Composition. pp. 43–54, (2005). [32] S. Dustdar and W. Schreiner, A survey on web services composition, International journal of web and grid services. (2005). [33] B. Yang and H. Garcia-Molina. Designing a super-peer network. In Proceedings of the 19th International Conference on Data Engineering, March 5-8, 2003, Bangalore, India, pp. 49–60. IEEE Computer Society, (2003). [34] M. Khambatti, K. D. Ryu, and P. Dasgupta. Peer-to-peer communities: Formation and discov- ery. In Proceedings of International Conference on Parallel and Distributed Computing Systems, PDCS 2002, November 4-6, 2002, Cambridge, USA, pp. 161–166, (2002). [35] I. Stoica, R. Morris, D. Karger, M. F. Kaashoek, and H. Balakrishnan. Chord: A scalable peer- to-peer lookup service for internet applications. In SIGCOMM ’01: Proceedings of the 2001 conference on Applications, technologies, architectures, and protocols for computer communications, pp. 149–160, New York, NY, USA, (2001). ACM. ISBN 1-58113-411- 8. doi: http://doi.acm.org/10.1145/383059.383071. [36] S. Ratnasamy, P. Francis, M. Handley, R. Karp, and S. Schenker. A scalable content-addressable network. In SIGCOMM ’01: Proceedings of the 2001 conference on Applications, tech-
- 44. 24 E. Abi-Lahoud, M. Savonnet, M.-N. Terrasse, M. Viviani, K. Yétongnon nologies, architectures, and protocols for computer communications, pp. 161–172, New York, NY, USA, (2001). ACM. ISBN 1-58113-411-8. doi: http://doi.acm.org/10.1145/ 383059.383072. [37] K. Oztoprak and G. B. Akar. Two-way/hybrid clustering architecture for peer to peer systems. In Proceedings of the International Conference on Internet and Web Applications and Services (ICIW 2007), May 13-19, 2007, Le Morne, Mauritius, pp. 11–16, (2007). [38] S. Ratnasamy, M. Handley, R. M. Karp, and S. Shenker. Topologically-aware overlay con- struction and server selection. In Proceedings of INFOCOM 2002: Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies., pp. 1190– 1199. IEEE Computer Society, (2002). [39] X. Y. Zhang, Q. Zhang, Z. Zhang, G. Song, and W. Zhu, A Construction of Locality-Aware Overlay Network: mOverlay and Its Performance, IEEE Journal on Selected Areas in Communications. 22(1), 18–28, (2004). [40] R. Zhang, A. R. Butt, and Y. C. Hu. Topology-Aware Peer-to-Peer On-demand Streaming. In Proceedings of NETWORKING 2005: Networking Technologies, Services, and Pro- tocols; Performance of Computer and Communication Networks; Mobile and Wireless Communication Systems, 4th International IFIP-TC6 Networking Conference, Water- loo, Canada, May 2-6, 2005, vol. 3462, Lecture Notes in Computer Science, pp. 1–14. Springer, (2005). [41] L. Ramaswamy, B. Gedik, and L. Liu, Connectivity based node clustering in decentralized peer- to-peer networks, p2p. 00, 66, (2003). doi: http://doi.ieeecomputersociety.org/10.1109/ PTP.2003.1231505. [42] W. Zheng, S. Zhang, Y. Ouyang, F. Makedon, and J. Ford. Node clustering based on link delay in p2p networks. In SAC ’05: Proceedings of the 2005 ACM symposium on Applied computing, pp. 744–749, New York, NY, USA, (2005). ACM. ISBN 1-58113-964-0. doi: http://doi.acm.org/10.1145/1066677.1066845. [43] World Wide Web Consortium (W3C). QoS for Web Services: Requirements and Possible Ap- proaches. http://www.w3c.or.kr/kr-office/TR/2003/ws-qos/. [44] N. Griffiths and M. Luck. Coalition formation through motivation and trust. In AAMAS ’03: Proceedings of the second international joint conference on Autonomous agents and mul- tiagent systems, pp. 17–24, New York, NY, USA, (2003). ACM. ISBN 1-58113-683-8. doi: http://doi.acm.org/10.1145/860575.860579. [45] Y. Wang and J. Vassileva. Trust-based community formation in peer-to-peer file sharing net- works. In WI ’04: Proceedings of the 2004 IEEE/WIC/ACM International Conference on Web Intelligence, pp. 341–348, Washington, DC, USA, (2004). IEEE Computer Society. ISBN 0-7695-2100-2. doi: http://dx.doi.org/10.1109/WI.2004.144. [46] K. Yetongnon, E. Abi-Lahoud, and M. Viviani. Service based application design in P2P overlay networks. In PDCCS-2008: ISCA 21st International Conference on Parallel and Dis- tributed Computing and Communication Systems, New Orleans, LA, USA (September, 2008).
- 45. Chapter 2 Complexity Analysis of Data Routing Algorithms in Extended Lucas Cube Networks Ernastuti and Ravi A. Salim Department of Computer Science, Gunadarma University, Jl. Margonda Raya 100, Depok, Indonesia E-mail: {ernas,ravi}@staff.gunadarma.ac.id We introduce a class of novel interconnection topologies called extended Lucas cube (ELC). The ELC is an induced subgraph of hypercube defined in terms of Fibonacci strings. This model is classified as a member of the Fibonacci cube family. ELC can serve as a framework for studying degraded hypercube due to faulty nodes or links. ELC maintains virtually almost all the desirable properties of hypercube. The focus of this paper is on the data communication aspects in the ELC. In this paper, we study data routing algorithms in the ELC, namely, unicast, broadcast and multicast algorithms. An important property of any message routing algorithm is to avoid deadlock. The unicast algorithm always succeeds in finding a path between them and ensures deadlock free in ELC. The time and traffic steps are used to measure the efficiency of routing algorithms. The unicast algorithm for ELC, which uses a Hamming distance path for any two nodes, is time and traffic optimal. The broadcast algorithm which employs the extended Lucas tree is traffic optimal and near time optimal. Two multicast algorithms are presented; they are based on an extended Lucas tree and a Hamiltonian cycle, respectively. 2.1 Introduction With advances in VLSI technology it has become feasible to build multicomputers consist- ing of hundreds or even thousands of processor nodes with local memory, which commu- nicate with each other over a fixed interconnection network. Essential conditions for the efficient use of such machines are routines for exchanging data between the processors. In view of the many network topologies and the multitude of communication patterns, it is not surprising that a rich body of theoretical and practical studies has been developed around the theme of communication. A. Gabillon et al., Web-Based Information Technologies and Distributed Systems, © 2010 Atlantis Press/World Scientific 25 Atlantis Ambient and Pervasive Intelligence 2, DOI 10.2991/978-94-91216-32-9_2,
- 46. 26 Ernastuti and Ravi A. Salim Many communication problems are special instances of the following (N, p,k1,k2-routing problem. N packets, each with its own source and destination, must be routed such that at most k1 packets are initially at any node, and at most k2 packets are finally at any node. The N packets reside on p nodes. Regular topologies offer the advantage that all nodes have a global knowledge of the network, allowing for simple routing and scheduling decisions. Special algorithms are also interesting if many of the packets that are sent or received by a node are the same. Of particular importance are the following basic operations: (1) A single node broadcast involves the transfer of a message from a particular node to all other network nodes; (2) A single node scatter is similar to single node broadcast except that different messages are broadcasted; (3) A multinode broadcast involves the simultaneous single node broadcast from all net- work nodes (there are different messages); (4) A total exchange (also called gossiping) is similar to multinode broadcast, except that all the packets sent are different; (5) A single node accumulate (also called gather) is the dual operation to single node scatter; and (6) A multinode accumulate is the dual operation to multinode broadcasting. Recently the hypercube has become a popular interconnection topology for parallel and dis- tributed processing. The popularity of the hypercube is due to its appealing properties such as logarithmic diameter and high bisection width, ease to embed other common structures, and many known efficient data communication schemes. Square and Palais in 1963 pro- posed a message passing multiprocessor computer with 2k processing nodes, in which each node is placed at the vertex of a k-dimensional hypercube and the edges of the hypercube are links between the processors. A problem with the hypercube topology is that the number of nodes in a system must be a power of 2. In practical terms, this is a severe restriction on the sizes of systems that can be built. This restriction can be overcome by using an incomplete hypercube, i.e., a hypercube missing certain of its nodes [6]. Unlike hypercubes, incomplete hypercubes can be constructed with any number of nodes. Incomplete hypercube network models which possess almost all attractive features of the hypercube were introduced by Wu [11], Hsu [5], Munarini [9] and Ernastuti [2] in 1993, 1997, 2002 and 2007, respectively namely Fibonacci Cube (FC), extended Fibonacci cube
- 47. Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 27 (EFC), Lucas Cube (LC), extended Lucas cube (ELC). These models are the induced sub- graphs of the hypercube that use about 1/5 fewer links than the comparable hypercube and its size does not increase as fast as the hypercube. They are restricted to be of certain sizes, i.e., they are Fibonacci numbers. Therefore, Sandi Klavzar [7] classified them as members of the Fibonacci cubes family. Though there are more Fibonacci numbers than numbers being power of 2, they do not fill the gap left by hypercubes very well. It can be shown that the node degrees in the FC, EFC, LC and ELC is a logarithmic function of the total number of nodes. This property provides improved fault tolerance over the incomplete hypercubes. They can be viewed as hypercubes with faulty nodes. They provides more choices of net- work size to the family of cube based structures. It has been also shown in [5, 11, 9, 4] that the FC, EFC, LC and ELC can be efficiently embedded many interesting structures such as hypercubes, linear arrays, rings and meshes. All these make them to be attractive interconnection topologies. The FC, EFC, LC and ELC have similar properties except in Hamiltonicity property [5, 11, 9, 3]. FC has Hamiltonian paths for every n, but only less than a third of them has Hamiltonian cycles. EFC and ELC have both Hamiltonian paths and Hamiltonian cy- cles for every n. As for LC, it has no Hamiltonian cycles at all for every n, albeit still has Hamiltonian paths in some n’s. The incomplete hypercube can be viewed as resulting from a complete hypercube after some nodes become faulty and the system is reconfigured [1]. Therefore, the FC, EFC, LC and ELC not only allow the construction of systems of arbitrary sizes, but also expose the nature of hypercube systems operating in a gracefully degraded mode. The incomplete hypercube with N nodes, where N could be any positive in- teger, is constructed in the same way as the hypercube. In other words, nodes are numbered from 0 to N − 1 and two nodes are linked if and only if their binary representations differ in exactly one bit. The incomplete hypercube suffers from a low degree of fault tolerance under certain condition. The reliability of data processing and data communication is very important in hypercube systems as in all parallel systems [1]. Efficient routing and broadcasting messages is a key issue to the performance of parallel and/or distributed systems. An important property of any message routing algorithm is to avoid deadlock [6]. The speed and the tolerance may be decreased if one or more processors or links become faulty [1]. In order to determine and avoid the faulty nodes and links in the data communication, there are many different kinds of methods to find the shortest paths between the source and the target nodes. In fact FC, EFC and LC have been proved possessing a simple routing algorithm [5, 11, 9].
- 48. 28 Ernastuti and Ravi A. Salim In this paper, the focus is to study the data communication aspects in the extended Lucas cubes (ELC). We use the basic operations of communication model for problem of a single node broadcast which involves the transfer of a message from a particular node to all other network nodes. Problem of a single node broadcast consists of three basic types of data routing, i.e., one-to-one (unicast), one-to-all (broadcast) and one-to-many (multicast). In this paper we apply data routing algorithms for ELC which refer to [10]. We show the unicast algorithm which uses a Hamming distance path for any two nodes for ELC. We also show the broadcast algorithm which employs the extended Lucas tree, and then we present two heuristic multicast algorithms based on an extended Lucas tree and a Hamiltonian cycle on ELC, respectively. To measure the efficiency of routing algorithms, the time and traffic steps are used. 2.2 Preliminaries and Notations We represent an interconnection topology by a graph G = (V,E), whereV (the set of nodes) denotes the processors and E (the set of edges) represents the communication links between processors; an edge is an unordered pair xy = {x,y} of distinct nodes of G. Sometimes, to avoid ambiguity, V and E are denoted by VG and EG. And we denote the number of nodes and edges of G by |VG| and |EG|. Definition 2.1. A path on a graph (also called a chain) is a sequence x1,x2,...,xn such that {x1,x2}, {x2,x3},...,{xn−1,xn}, are edges of the graph and the xi are distinct. A closed path (x1,x2,...,xn,x1) on a graph is called a graph cycle or circuit. Definition 2.2. For x, y ∈ VG, dG(x,y) or d(x,y), denotes the length of a shortest path (a path with the least number of edges) in G from x to y. Let {0, 1}n denote the set of length n binary strings. Definition 2.3. The Hamming distance between two binary strings x, y ∈ {0, 1}n denoted H(x,y), is the number of bits where x and y differ. Definition 2.4. The Hypercube of dimension n, denoted by Q(n), is the graph, where the set of labels of nodes is {0, 1}n and two nodes x and y are adjacent if and only if their labels differ in exactly one bit (in other words H(x,y) = 1). Fig. 2.1 shows examples of Q(n), for n = 1, 2, 3, 4 respectively.
- 49. Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 29 Figure 2.1 Hypercube of dimension 1, 2, 3, 4: Q(1), Q(2), Q(3) and Q(4) Definition 2.5. The Hamming distance between two binary strings x, y ∈ {0, 1}n denoted H(x,y) shows the length of shortest path between node x and node y. Definition 2.6. The Fibonacci numbers form a sequence of positive integers fn, where f1 = 1, f2 = 1 and fn = fn−1 + fn−2, for n 2. Definition 2.7. A Fibonacci string of length n is a binary string a1a2 ...an which belongs to {0, 1}n with aiai+1 = 0, 1 ⩽ i n. In other words, a Fibonacci string is a binary string of length n with no two consecutive ones. It is easy to see that the number of Fibonacci strings of length n is the (n + 2) Fibonacci number (this connects Definition 2.6 and 2.7). The definition of FC, EFC, LC and ELC are based upon Fibonacci strings and the Hamming distance. 2.3 Graph Models of Fibonacci Cube Family FC, EFC and LC topologies use the Fibonacci sequence; however the initial conditions among them may differ from the initial conditions of the Fibonacci sequence. In this section we show the differences. The symbol · denotes a concatenation operation; for example, 01 · {0,1} = {010,011} and 01 · { } = {01}. The FC, EFC and LC can be respectively described as below. Definition 2.8 (Fibonacci cube [5]). For n ⩾ 0, the Fibonacci cube FC(n) = (VFC(n),EFC(n)) is defined as follows: VFC(n), the set of labels of nodes in FC(n), is recursively defined as VFC(n) = ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ ∅ if n = 0 {λ} if n = 1, 2 {0, 1} if n = 3 0 ·VFC(n −1)∪10 ·VFC(n −2) if n 3
- 50. 30 Ernastuti and Ravi A. Salim Two nodes in VFC(n) are connected by an edge in EFC(n) if and only if their labels differ exactly in one position. An FC(n) contains two disjoint subgraphs that are isomorphic to FC(n−1) and FC(n−2) [5]. Fig. 2.2 shows examples of FC(n), with n = 3, 4, 5, 6 respectively. Property 2.1 ([5]). For any n ⩾ 3, |VFC(n) = fn, where fn is the nth Fibonacci number. Figure 2.2 Fibonacci cube (a) FC(3), (b) FC(4), (c) FC(5), (d) FC(6) Definition 2.9 (Extended Fibonacci cube [11]). For n ⩾ 0, the extended Fibonacci cube EFC(n) = (VEFC(n),EEFC(n)) is defined as follows: VEFC(n), the set of labels of nodes in EFC(n), is recursively defined as VEFC(n) = ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ ∅ if n = 0 {λ} if n = 1, 2 {0, 1} if n = 3 {00, 10, 11, 01} if n = 4 0 ·VEFC(n −1)∪10 ·VEFC(n −2) if n 4 Two nodes in VEFC(n) are connected by an edge in EEFC(n) if and only if their labels differ exactly in one position. An EFC(n) contains two disjoint subgraphs that are isomorphic to EFC(n − 1) and EFC(n −2) [11]. Fig. 2.3 shows examples of EFC(n), with n = 3, 4, 5, 6 respectively. Property 2.2 ([7]). The number of nodes of EFC(n) is 2 fn−1, where fn is the nth Fibonacci number. Definition 2.10. An extended Fibonacci tree T1(n) of EFC(n) is defined as follows: (Base) T1(3) and T1(4) are defined as shown in Fig. 2.4a and 2.4b. Basically, T1(3) is EFC(3) with node 0 being the root and T1(4) is an EFC(4) rooted at node 00 after removing the link
- 51. Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 31 Figure 2.3 Extended Fibonacci cube (a) EFC(3), (b) EFC(4), (c) EFC(5), (d) EFC(6) connecting nodes 01 and 11. (Recursion) T1(n) (n 4) consists of T1(n −1) and T1(n −2) by connecting the root of T1(n −2) as a child of the root of T1(n −1). Suppose T1(n) also denotes the set of nodes in T1(n), then T1(n) = 0 ·T1(n −1)∪10 ·T1(n −2). Fig. 2.4 shows examples of the extended Fibonacci tree T1(n), for n = 3, 4, 5, 6 respectively. Figure 2.4 Extended Fibonacci tree (a) T1(3), (b) T1(4), (c) T1(5), (d) T1(6) Definition 2.11 (Lucas cube [9]). For n ⩾ 0, the Lucas cube LC(n) = (VLC(n),ELC(n)) is defined as follows: VLC(n), the set of labels of nodes in LC(n), is recursively defined as VLC(n) = ⎧ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎩ ∅ if n = 0 {λ} if n = 1, 2 {0, 1} if n = 3 0 ·VFC(n −1)∪10 ·VFC(n −3)·0 if n 3 Two nodes in VLC(n) are connected by an edge in ELC(n) if and only if their labels differ exactly in one position. For any n ⩾ 0, The FC(n), EFC(n) and LC(n) are induced subgraphs of Q(n−2) [5, 11, 9].
- 52. 32 Ernastuti and Ravi A. Salim 2.4 Extended Lucas Cube (ELC) The ELC is defined on the same way of LC recurrence by using the EFC as its initial condition. The following definition gives a recursive definition for ELC. Definition 2.12 (Extended Lucas cube [3, 2]). For n ⩾ 0, the extended Lucas cube ELC(n) = (VELC(n),EELC(n)) is defined as follows: VELC(n), the set of labels of nodes in ELC(n), is recursively defined as VELC(n) = ⎧ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ ∅ if n = 0 {λ} if n = 1, 2 {0, 1} if n = 3 {00, 10, 11, 01} if n = 4 0 ·VEFC(n −1)∪10 ·VEFC(n −3)·0 if n 4 Two nodes in VELC(n) are connected by an edge in EELC(n) if and only if their labels differ exactly in one position. Definition 2.12 says thatVELC(n) ⊆ {0, 1}n−2, n ⩾ 2, and so ELC(n) is an induced subgraph of Q(n −2). Fig. 2.5 shows examples of ELC(n), with n = 3, 4, 5, 6 respectively. Figure 2.5 Extended Lucas cube (a) ELC(3), (b) ELC(4), (c) ELC(5), (d) ELC(6) The properties of the extended Lucas cube (ELC) are given below. Refer to [2] for more detail and proofs to these properties. Property 2.3. For n ⩾ 3, ELC(n) contains two disjoint subgraphs that are isomorphic to ELC(n −1) and ELC(n −3), respectively. Property 2.4. For n ⩾ 4, n = 5, ELC(n) is a Hamiltonian graph. Property 2.5. There exists a Hamming distance path between any two nodes in ELC.
- 53. Complexity Analysis of Data Routing Algorithms in ExtendedLucas Cube Networks 33 Property 2.6. The Hamming distance path between any two nodes in ELC is a shortest path. [2] has proved that ELC(n) contains two subgraphs which are isomorphic to EFC(n − 1) and EFC(n −3), respectively. There are exactly fn−3 edges linking those two subgraphs. Property 2.7. For any n ⩾ 5, in ELC(n), there are exactly fn−3 edges linking subgraph induced by 0 ·VEFC(n −1) to subgraph induced by 10 ·VEFC(n −1)·0. Property 2.8. For any n ⩾ 3, ELC(n) is a connected graph. Property 2.9. Diameter of ELC(n) is n −2, for n ⩾ 3. Property 2.10. The node degree of a node in ELC(n), n ⩾ 3, is between n −3 3 and n−2, except for n = 4, the node degree is n −2. Property 2.11. For n ⩾ 3, |VELC(n)| = |VEFC(n −1)|+|VEFC(n −3)|. Property 2.12. For any n ⩾ 5, the number of nodes of ELC(n) is 2 fn−2 +2 fn−4. Table 2.1 shows the number of nodes of hypercube, FC(n), EFC(n), LC(n) and ELC(n) for 3 ⩽ n ⩽ 12. Definition 2.13. An extended Lucas tree T2(n) of ELC(n) is defined as follows: (Base) T2(3) and T2(4) are defined as shown in Fig. 2.6a and 2.6b. Basically, T2(3) is ELC(3) with node 0 being the root and T2(4) is an ELC(4) rooted at node 00. (Recursion) T2(n) (n ⩾ 4) consists of T1(n − 1) and T1(n − 3) by connecting the root of T1(n − 3) as a child of the root of T1(n−1). Suppose T2(n) also denotes the set of nodes in T2(n), then T2(n) = 0 ·T1(n −1)∪10 ·T1(n −3)·0. Property 2.13. Extended Lucas Tree T2(n) is a spanning tree of T2(n). Property 2.14. T2(n) contains two disjoint subtrees that are isomorphic to T1(n − 1) and T1(n −3), respectively. Property 2.15. For any n ⩾ 3, the span of T2(n) is n −2. Property 2.16. For any n ⩾ 3, the height of T2(n) is n −2 2 . Fig. 2.6 shows examples of the extended Lucas tree T2(n), for n = 1, 2, 3, 4 respectively. Property 2.17. In T2(n), the children of the root are dimension ordered, i.e. the ith child of the root is the neighbor of the root on the ith dimension.
- 54. 34 Ernastuti and Ravi A. Salim Figure 2.6 Extended Lucas tree (a) T2(3), (b) T2(4), (c) T2(5), (d) T2(6) Table 2.1 Table of the number of nodes of hypercube, FC, EFC, LC and ELC n VQ(n−2) VFC(n) VEFC(n) VLC(n) VELC(n) 3 2 2 2 2 2 4 4 3 4 3 4 5 8 5 6 4 5 6 16 8 10 7 8 7 32 13 16 11 14 8 64 21 26 18 22 9 128 34 42 29 36 10 256 55 68 47 58 11 512 89 110 76 94 12 1024 144 178 123 152 Property 2.18. The pre-order of T2(n) is the same as the order by the binary values of node addresses. 2.5 Data Routing Algorithms in ELC Data communication is the delivery of message from the source to the destination. In gen- eral, there are four types of communications based on the sets of the source and destination nodes, one-to-one, one-to-all, one-to-many and many-to-many. Unicast (one-to-one) is the term used to describe communication where a piece of infor- mation (message) is sent from a single source node to a single destination node. In this case there is just one sender, and one receiver. Broadcast (one-to-all) is the term used to describe communication where a piece of information (message) is sent from a single source node to all the other nodes. In this case there is just one sender, but the information is sent to all connected receivers. Multicast (one-to-many/ many-to-many) is the term used to describe communication where a piece of information (message) is sent from one single source node
- 55. Another Random Document on Scribd Without Any Related Topics
- 56. who have been in the trade a lifetime will recall the details of almost every voyage they have made—the time of starting, the shifts of wind, the margin of time by which they saved their tide, what they saw on the way, and a dozen other things—never confusing one passage with another. When you sail by bargees or smacksmen at anchor you behold them apparently staring aimlessly on to the sea or into the sky; but they are watching. Perhaps they seem to be looking the other way, but they have marked you pass and noticed, it may be, that your topping lift is too taut. This or any other detail is duly entered in the unwritten log of their memories. On shore they take their leisure on the quay, walking up and down, never more than a few steps each way, with eyes always on the anchorage. The arrival of a stranger, the way he anchors, the coming and going of dinghies, the manner in which they are brought alongside—everything is noted. Now, the chief object of interest in the gear of the Will Arding was a new kedge anchor. To men accustomed to anchor near the shore and in very narrow swatchways nothing is more important than their ground tackle. They spend more anxious thought on that than on anything else. My new anchor was lying on the quay, and I could hear the comments of every passer by. I was flattered by an accumulation of approval. Sometimes I was below, and did not know who was speaking; nor did it much matter, since the language of all was interchangeable. I would simply hear a voice; and soon another voice would be saying the same thing over again. Imagine a succession of observations like this: First Voice: ‘Yes, yes; that’s a good anchor, that is. As I was a sayin’ to Jim this mornin’, “That’s got good flues, that has, and a good stock. I lay she ’on’t never drag that,” I says, “if that git aholt in good houldin’ graound. No more she ’on’t faoul that. That’ll hould she in worse weather than what they’ll ever want to be aout in,” I says. “Then agin, that’s a good anchor for layin’ aout, for that ain’t a heavy anchor to handle in a bo’t,” I says. “None the more for that, she ’on’t never drag that. The chap what made that anchor knaowd what he was abaout.”’
- 57. Second Voice: ‘That’s a wonnerful good anchor, that is. That ’on’t never drag that if they let that goo in good houldin’ graound. I allus did like an anchor long in the stock, same as that. Yes, yes; that’ll hould she. That ain’t a heavy anchor for same as layin’ off in a bo’t, whereas them heavy anchors is wonnerful ill convenient. Yes, yes; they’ve got a good anchor there; that was made at Leigh, that was, and wonnerful good anchors that smith allus did make.’ Third Voice: ‘What do I think in it? I don’t want to think nawthen abaout that. I knaow that’s a good anchor. She ’on’t never drag that, do, that’ll hev to be wonnerful poor houldin’ graound. That anchor’s got good flues, that has, and she ’on’t never drag that nit faoul it. They’ll want to be in harbour time that anchor ’on’t hould she. That’s long in the stock, that is, but none the more for that that ain’t a heavy anchor, and yaou can lay that aout in a bit of a sea when maybe a heavier un ’ould be too much for yer.’ The next day the Mate and the elder boy returned, and the barge was christened with a new name. Will Arding, no doubt, had had some sufficient meaning for the late owner, but for us it meant nothing, and we had decided to call the barge Ark Royal. Before the christening we moved from the quay into midstream. The warps ashore were cast off, and the clank, clank, clank, of the windlass sounded like the music of other worlds calling. We slowly hove off the barge until her stern swung round and she rode free to the flood-tide and the east wind. Sam Prawle was on board, as I had engaged him to come for our first cruise in order that I might learn the handling of a barge under a good instructor. We could not start till high water, because the wind was up river. Meanwhile, the christening was performed. Several smacksmen came off in their boats for the ceremony. A bottle of champagne, made fast to the jib topsail halyards, was flung well outboard, and came back on to the barge’s bluff bows with a crash and an explosion of foam as the Mate said: ‘In the name of all good luck I christen you Ark Royal!’ Everyone cheered; other champagne (not the christening brand) was handed round, and we all drank success and long life and happiness to
- 58. one another and the ship. The Royal Cruising Club burgee was hoisted to the truck and the Blue Ensign at the mizzen peak. Sam stowed the wine-glasses in their racks below; the good-byes were said; the smackies clambered over the side, sorted themselves into the cluster of dinghies astern, and lay on their oars to watch the start. The tide was on the turn, the great topsail flacked in the wind, the brails were let go, and Sam and I sweated the mainsheet home and set the mizzen.
- 59. The Ark Royal She was feeling the ebb now, and she sheered first one way and then the other, gently tugging at her anchor as we hoisted the foresail and made the bowline fast to port. Once more the clank, clank, of the windlass; the short scope of the bower anchor came home sweetly, and the Ark Royal was free. I left Sam to get the anchor right up and flew aft to the wheel as she slowly gathered way. We were off! Good-bye to the land and houses and rates and by-laws! We believed that we were entering on a better way of life. We have since made sure of it. I think of that first sail still. The newness to us of the Ark Royal’s great size; her height above the water; the grand sweep she took as she came about; the march from the wheel to the leeside to peer forward in bargee’s style to see whether there was anything in our way to leeward; the size of the wheel itself, and the many turns wanted to put the helm down or up, filled us with importance and pride as we tacked down the river. If you would know what my feelings were then you must think of your first boundary to square leg, your first salmon, your first gun, your first stone wall with hounds running fast. That night we anchored at the mouth of the river, and when the sails were stowed and the riding light had been hoisted, we ate our first dinner on board and tucked our elder boy into his bunk for the first time. Then beneath the stars, rocking gently on a scarcely perceptible easterly swell, we walked our decks in the flood-tide of happiness. ‘None of our relations know where we are or where we are going to,’ said the Mate. ‘Here we are now, and to-morrow, perhaps, we shall get to Mersea Island and pick up Margaret and Inky, and then we shall be complete. Is it real? Is it true?’ We sat on deck very late, too much occupied with the pleasure of existing to yield to sleep. The sky was continually changing as snowy clouds drifted across it. In the distance the Swin Middle light flared up
- 60. like a bonfire every fifteen seconds. Here and there the lights of barges drooped tremulous threads of gold on the water. Sam Prawle was invited aft; and regarding us now as freemen of the barge profession, he enlarged upon the advantages of barging (comparing it with the sport of yachting, which he seemed to think we had abandoned) with a confidential note in his voice that we had not precisely detected before. But his opinions on these weighty matters deserve a chapter to themselves.
- 61. CHAPTER XII ‘Vous êtes tous les deux ténébreux et discrets: Homme, nul n’a sondé le fond de tes abîmes, O mer, nul ne connaît tes richesses intimes, Tant vous êtes jaloux de garder vos secrets!’ Seated on the after cabin-top near the wheel, Sam Prawle made known to us the arcana of barging. The comparison with yachting was to the disadvantage of yachting, and we felt that he would not have ventured to take this line had we still owned the Playmate. On the other hand, we were gratified at being treated with frankness as members of his profession. ‘I don’t reckon,’ said Sam Prawle, ‘there ain’t nawthen as good as bargin’, same as on the water, my meanin’ is. Ye see, yaou gets home fairly frequent, yaou ain’t got no long sea-passages to make, yaou can see a bit o’ life in the taowns, and ef yaou’ve got a good little ould barge and freights is anyways good ye can make a tidy bit o’ money. ‘Then agin, in respect o’ livin’, most all barges carries a gun, and there’s some I could name as carries oyster drudges; then there’s a bit o’ fishin’ to be done, and accordin’ to where yaou’re brought up there may be winkles, or mussels, or cockles, and, as I says, chance time a few oysters; so my meanin’ is the livin’ is good. ‘A course that don’t do for it to be knaown ye carries a drudge no more than that do to be seen pickin’ up oysters nit winkles in some places, same as on the Corporation’s graounds in the Maldon River. But outside them graounds that does no detriment. I dessay yaou remember some time back abaout they chaps what was caught pickin’ up winkles in the Maldon River. Well, the judge give it agin them, for a course the Corporation has all the fishin’ rights above them beacons. But the most amusingest part was, they chaps’ lawyer tried to make
- 62. aout a winkle warn’t a fish, but a wild animal. Yes, yes; they lost right enough. ‘Us allus used to live wonnerful well on the ould Kate, for I had a mate, Bill Summers, who was a masterpiece at shoot’n’. He were suthen strorng, he were, and had masterous great limbs on ’im, but none the more for that he were a wonnerful easy-spoken chap. I’ve knaowed he caught a many times by same as keepers and that, but he allus had some excuse or spoke ’em fair. Leastways, he den’t never git into trouble. ‘I remember one November day there’d bin a heavy dag in the fore part o’ the day which cleared off towards the afternoon, and Bill went ashore after a hare or whatever he could git daown on they ould mashes away to the eastward there. A wonnerful lonely place that is— no housen nor nawthen but they great ould mashes. A course Bill den’t reckon there’d be anyone a lookin’ after the shootin’ daown there, but there were. But as I was a tellin’ yer, Bill most allus knaowed what to say to such as they. Well, just afore that come dark, about flight time, I raowed the boat ashore to the edge o’ the mud on the lookaout for Bill. I waited some time, and that grew darker and darker, and them watery birds and curlew kep’ all on a callin’, and one o’ they ould frank- herons come a flappin’ overhead, and that fared wonnerful an’ lonesome. ‘Well, I was jist a wonderin’ whether I hadn’t better goo and look for Bill in case he’d got stuck in one o’ they fleets what run acrost mashes, or had come to some hurt, for a man might lay aout there days and weeks afore anyone might hap to find ’im. Then I heard suthen and sees Bill a comin’ suthen fast along the top o’ the sea-wall with another chap a comin’ arter ’im. “Ullo,” I thinks, “Bill’s in trouble,” so I gives a whistle, and Bill answers and comes straight on daown the mud towards the bo’t with his gun in one hand and an ould hare or suthen in the other. When he gits half-way daown the mud Bill turns raound to the chap a follerin’ and says, “Do yaou ever read the noospapers, mate?”
- 63. ‘The chap, he den’t say nawthen, so Bill stops and ’as a look at ’is gun, and then he says agin werry slow, “Funny things you reads of ’appenin’ in the noospapers.” ‘Well, that chap den’t fare to come no further, and Bill finishes ’is walk daown the mud alone. Wonnerful easy-spoken chap, ’e was. Yes, yes; us allus had good livin’ on the Kate. ‘Then agin, same as summer-time, maybe yaou’ve got a fair freight, or yaou’re doin’ a bit o’ cotcheling, and yaou’re a layin’ up some snug creek, and the tides ain’t just right for gittin’ away, and yaou has to wait three or faour days. Well, that’s wonnerful comfortable, that is, specially ef there’s a bit of a village handy. Or same as layin’ wind- baound winter-time, maybe twenty barges all together—and I remember sixty-two layin’ wind-baound at the mouth o’ the Burnham River once’t—well, that’ll be a rum ’un if there ain’t a bit o’ jollification goin’ on aboard some o’ they. Yes, yes; I allus says bargin’ is what ye likes to make it. ‘What other craft can a man take his missus in—leastways, ef he has a mind to? They what ain’t got little ’uns often takes their wives with ’em, and summer-time they can often manage without a mate in same as ninety-ton barges. A course, that’s a bit awk’ard ef ye gits into trouble, for a woman can’t do what a man can, and a man can’t allus say what he wants to ef he has the missus with him. ‘But that’s true, women’s wonnerful artful, and I’ve knaowed a woman say suthen more better than what a man could. When ould Ted Wetherby—a wonnerful hard-swearin’ man—took his missus with him, they was nearly run daown by a torpedo bo’t in the Medway. That young lootenant in charge pitched into Ted suthen cruel, but Ted he den’t say nawthen till that young chap was abaout in the middle of what ’e ’ad to say, and then ’e jist up and says, “Ush! Ladies at the hellum!” And then the lootenant turns on Ted’s missus, and tells she jist what he thought about Ted and the barge. Ted’s missus den’t say nawthen neither till they was jist sheerin’ off, and then she says, “I don’t take no more notus o’ what yaou say than ef ye ain’t never
- 64. spoke.” Bill tould me he reckoned that lootenant were more wild than ef Bill ’ad spoke hisself. ‘Then agin, a skipper of a barge is most all the time his own master in a manner o’ speakin’. A course, some says yachtin’ is easier, and maybe it is, but I don’t hould with it. I’ve met scores o’ yacht skippers and had many a yarn along o’ they, but I’d rather be skipper of a little ould barge than any yacht afloat. My cousin, Seth Smith, is skipper of a yacht, and he’s tould me some o’ the wrinkles o’ yachtin’. ‘From what I can ’ear of it, there’s owners and owners. Accordin’ to some, they what don’t knaow nawthen fare to be the best kind to be with. Leastways, that’s a wonnerful thing haow long a yacht will lay off a place the skipper and crew likes. I remember one beautiful little wessel a layin’ off the same blessed ould place week after week, so I ast a chap I knaowed if she den’t never git under way. “Well,” ’e says, “yaou see, the owner, he don’t knaow nawthen, and the skipper and crew belongs ’ere. Chance time they do get under way, but we most allus says o’ she ’ef there ain’t enough wind to blaow a match aout there ain’t enough wind for she to muster, and ef there’s enough wind to blaow a match aout that’s too much for she, as the sayin’ is.” ‘But there’s owners what sails their own wessels, and Seth says as haow they is good enough to be along with, for ef they gits into trouble they gits into trouble, and that ain’t nawthen to do with the crew. ‘But they owners what knaows a little is the worst, because they thinks they knaows everything, in a manner o’ speakin’, and the skipper has to be wonnerful careful. Yaou see, the trouble lays along o’ the steerin’. A course, most anyone can steer, though they don’t git the best aout of a wessel, but same as owners an’ they allus fare to reckon that steerin’ is everything, which a course it ain’t. Seth has tould me a score o’ times, he has, “Sam,” he says, “that’s a strain on a man, that is, for he’s got to keep all on a watchin’ his owner to see he keeps the wessel full or don’t gybe she, or one thing an’ another. Naow same as tackin’ up this ’ere little ould river,” he says, “or standin’ into shaoal water, ye just says to me comfortable like, ‘Shove the ould gal round,’ whereas
- 65. my meanin’ is that ’on’t do for a yacht skipper to say that to his owner. No, no; that ’on’t do; he’s got to goo careful like. Maybe he’ll say, ‘What do you think abaout comin’ abaout sir?’ Then maybe—if there ain’t no visitors aboard—the owner’ll say, ‘Let ’er come.’ Then agin, maybe there’s visitors aboard, and the owner ’e takes a look raound and says, ‘In another length,’ or suthen o’ that.” ‘But ef the skipper’s bearin’ a hand with suthen, or for one thing or another he leaves that a bit late, so as he ain’t got time to ask the owner what e’ thinks and let him have his look raound so that fare as haow he’s in charge, but jist says, “Shove her round,” quick like, then the owner ain’t over and above pleased—especially if there’s visitors aboard, as I was a sayin’. That’s ill convenient, that is, for ef she don’t come raound quick enough she’ll take the graound, and then the skipper’s got to say a hill has graowed up or a landmark’s bin cut daown or suthen, and kaidge she off too; and a course, same as on the ebb, that’s a hundred to one she ’on’t shift till she fleet next tide. Yes, yes; a skipper’s got to be wonnerful forehanded as well as careful what ’e says. ‘I remember a friend o’ mine, Jem Selby, goin’ along of a gent who was wonnerful praoud o’ his cruises, what ’e did without a skipper. He on’y took Jem, he said, cos Jem were a deep-water man and hadn’t never been in a yacht afore, but on’y in same as barques and ships and wessels similar-same to that, and ’e wanted a man just to cook and put him ashore. Well, this gent and Jem brought the little yacht—I can’t remember her name—from Lowestoft daown to Falmouth, and the gent was wonnerful praoud o’ hisself, as they’d been aout in some tidy breezes. He was a tellin’ of his friends at Falmouth all abaout his adventures, and the gales o’ wind they had come through, when he turns to Jem, who was standin’ by, and says, “What do yaou say to goin’ raound Land’s End to-morrer, Jem?” “Well, I don’t knaow, sir,” says Jem; “yaou see, we’re a gettin’ near the sea now.” Maybe it were that, maybe it warn’t, but ’e den’t ast Jem to sail along o’ he next season. ‘Well, there yaou are now. Ye can’t do nawthen and ye can’t say nawthen. No, no; from what I can ’ear of it and from what I can see of
- 66. it, yachtin’ ain’t in the same street as bargin’, as the sayin’ is. Let alone, some o’ they chaps never does a hand’s turn o’ work from one week to another ’cept maybe polish a bit o’ brass work. ‘Seth says as haow that ain’t a bad job to be in charge of a little yacht with a party o’ young chaps, same as on their holiday. Young chaps, same as they, never drinks without the skipper, and a course they most allus lives well, so the skipper do too. Then agin, yaou see they likes to do all the work, and the skipper just puggles abaout like and tells they what to do, though a course they wants lookin’ arter none the more for that. Maybe on dewy nights the skipper ’as to goo raound quiet like and ease up the halyards, for young chaps is all for havin’ everything smart and taut; but that ain’t nawthen, and he can most allus do that while they has their supper. ‘From what I see of it myself, I reckon young chaps same as they is a bit troublesome goin’ into harbour. I remember seein’ a party o’ faour come into Lowestoft in a little yacht—a doddy little thing, she were— with an ould fellow in charge. The Lord Nelson was just startin’ for Yarmouth, so they couldn’t berth until she’d gone, and as I happed to be standin’ by I made fast the lines the ould chap thraowed on the pier. Well, the band was a playin’ and the pier crowded with gals a watchin’ the yachts in the harbour, and they young chaps den’t fare to be able to keep quiet like with them gals a lookin’ on, and kep’ all on worritin’ the ould chap to knaow ef they hadn’t better give a pull on this or a pull on t’other. Then I seed the artful ould chap give one on ’em the headrope to hould and another the starn rope—though they might just as well a bin made fast—and another he give a fender to, and t’other one, what was the most worritsome o’ the lot, ’e took and made fast the jib sheets raound the bitts and tould he to pull on that. And he did. Lor’, that did make me laugh suthen. ‘Then agin, some o’ they young ’uns hears things what they den’t ought to. I remember young Abe Putwain, who used to sail along of a wonnerful larned ould gent what was always a lookin’ at things he got out o’ the water with one o’ they microscopes—a master great thing that were, accord’ to Abe. Well, this ould party and his friends was most allus argyin’ abaout suthen, and a course Abe could hear they
- 67. through the fo’c’sle door. Abe was the most reg’lar chapel man I ever knaowed, and used allus to hould the plate by the door every Sunday till he took up along this larned gent what I’m a talkin’ abaout. Just abaout Christmas my mate left to take a skipper’s job, so bein’ at home I says to Abe, who I ain’t seen for some bit, “Will you come, mate, along o’ me, as yaour bo’t’s laid up?” So he come as mate, and one day, when we was sailing daown past the Naze and had just opened up Harwich Church, I says, “Well, mate, there’s the ould church!” I says, meanin’ the landmark. “Oh,” ’e says, scornful like. “You don’t ’ould with them idle superstitions, do yer?” he says. Well, that warn’t no use argyin’ with he, for he ain’t never bin to chapel since, and that’s what come o’ yachtin’, I reckon.’
- 68. CHAPTER XIII ‘Here are our thoughts—voyagers’ thoughts, Here not the land, firm land, alone appears, may then by them be said; The sky o’erarches here—we feel the undulating deck beneath our feet, We feel the long pulsation—ebb and flow of endless motion; The tones of unseen mystery—the vague and vast suggestions of the briny world—the liquid-flowing syllables.’ The riding light was already garish in the early sunshine when we turned out the next morning. The fragrance of the breeze coming in faint puffs off the land, the clean taste of the air, the cries of the sea birds, and the tender haze that overhung the land, set all our senses tingling. Yet what a creature is man! As we stood by the main rigging there came wafted aft to us from the forehatch the bubbling sound and the smell of frying bacon, and we could scarcely endure the delay of staying to wash down the decks, though that was a duty to be performed before hunger might be satisfied honourably. We got under way soon after breakfast, but the wind was fluky and we drifted rather than sailed. About low water we anchored in a clock calm to wait for the easterly breeze which we knew would come later, for the gossamers hung on the rigging. In the afternoon the wind duly ‘shot up at east,’ as the fishermen say, and we fetched over the Dengie flats, opened the Blackwater, and bore away for Mersea Island to pick up the other children. We anchored in the Deeps, for there was no room for such a large vessel as ours in our old haunts up the creeks, but before the anchor was down two small figures in white came running down King’s Hard. Inky and Margaret had been watching for us. We soon had the sailing
- 69. dinghy going off for them. How pleased they were, how excited about their cabins, how astonished at finding their toys ready for them! At last, then, our scheme was complete. The family was reassembled under a new roof, and that roof was a deck. We met several sailing friends at West Mersea, and found our old yacht, the Playmate, from whose owners we heard an account of their first trip to Mersea. Off the entrance they hailed the man on board the watchboat, to ask the way into the quarters. The watchman, who had known the Playmate for years, and had seen her going in and out scores of times, answered the question in the spirit in which he supposed it had been asked. He had not heard that the vessel had changed hands. ‘Go on. Yaou knaow,’ he shouted back. ‘No, we don’t,’ bawled the new owners. ‘Go on. Yaou knaow,’ he repeated, as the Playmate forged on. ‘No, we don’t,’ yelled the new owners, becoming nervous of running aground. ‘Yaou let the ould girl goo herself, then. She knaow the way in!’ was the last they heard. During our short cruise we found out how best to arrange everything on board so as to avoid breakages in a sea. Our furniture, of course, had not been specially made for a ship; some of it had already been screwed to the walls or bulkheads; the rest of it could be quickly wedged. The shelves were all fitted with ledges, so that china and silver had only to be laid flat behind the ledges. On deck we hung thin boards over the windows, as these might easily be broken. At Osea Island in the Blackwater we took in eight hundred gallons of water. We then visited Heybridge, Brightlingsea, and Wivenhoe, and still left ourselves ample time to make the passage to Newcliff and settle down comfortably before the boys were due at their school.
- 70. To revisit the Essex sea-marshes is always to discover something new. The dim low land may be called dreary compared with the more vivacious Solent, but when the spell of this Dutch-like scenery has been laid on you it has touched your heart for ever. Not all people who are in love with Essex have always been so. The charms of the county inland, as well as on the coast, have to be discovered gradually, because they are widely spread. Essex has no cathedral which gathers up the interest to one point. Yet its houses are an epitome of its history and character; they look as though they were part of the landscape, as though they had grown up with the trees. Some houses in Essex—farmhouses and inns—often welcome you with a clean white face, but the complexion of a whole village seen far off is nearly always red, and a thin spire generally tapers above the roofs. Churches and houses alike were built with the materials which were ready to hand. There is much timber in the building, because Essex has few quarries. In hundreds of churches, too, you may see the relics of the Roman occupation. The Roman bricks are worked into the lower parts of the walls; flint commonly comes above the brick, and stout timbers are used not only for the roof, but in the whole construction. Sometimes the spire is made entirely of wood, and there is surely something beautiful and touching in the exaltation to this use of the characteristic material of the county. When a beam was wanted for a house, or a roof for a church, chestnut was the wood, no doubt because of the belief that no insect takes kindly to it. The great building age of what is now rural Essex must have come immediately after the suppression of the monasteries, and you can hardly go into an Essex village without finding a Tudor house. If it be a manor-house, it may have a moat or a monkish fishpond; and perhaps the pigeon tower, which dates from the times when the lord of the manor had his rights of pigeonry, is still standing. The old inns have a spaciousness which informs you of the well-being of agricultural Essex when they were built. Where the land is good there the inns are good also; where the land is poor the inns are built on niggard lines. You can come across Essex villages—such as the Rodings, the Lavers, and the Easters—which for remoteness of air and unsophistication
- 71. could not be matched except in counties so distant from London as Cornwall and Cumberland. Certainly Essex has no great hills, even as it has no great buildings. But the value of hills is relative. From many places in Essex only about sixty feet above the sea there are wide views, and you may gaze upon the Kentish coast thirty miles away on the other side of the Thames. The secret of the Essex coast is the illusion of immensity. The dome of sky is scarcely interrupted by the small frettings of land and wood along the edges. In this vast atmospheric theatre a change of weather may be seen at almost any point of the compass planning its tactics on a clear hard line of horizon, and thence swinging up the sky, showing the soft white flags of peace or the threatening front of a battle formation. One even has an important sense of the monstrous nearness of natural forces when the ‘inverted bowl’ is filled with a dark low-flying scud that seems to be crushing down on you in a kind of personal assault. Men who have become captivated by the marshes have been able to measure the gradual and unconscious change in their feelings about hills and flat lands by a visit to some such spot as the Italian Lakes. The beauty of the lakes has always to be admitted—the purity of the water, the affluence of the colour, the abrupt fall of the hills to the water, the sweetness of the glinting villages perched high up as though resting in a long and difficult climb to the sky. But at the end of a week the visitor may have found himself insisting on these beauties; he has felt that the sense of them is slipping away. He who needs to argue with himself is losing ground. He becomes unreasonably conscious that the water is imprisoned, and does not lead to the sea round the distant headland; that the sky is filched away; and that the winds are false, being misdirected by the hills and simply blowing up or down a long corridor, so that Nature is frustrated in these coddled and enchanted haunts. In shallow estuaries like those of Essex the tides have necessarily to be studied more carefully than in deep waters. The ebb tide runs faster than the flood; for the ebb is hurried seawards, pressed on its flanks as it goes, by the weight of water that pours off the flats from either
- 72. side of the channel. The flood comes in from the sea like a cautious explorer. It is as though it could afford to be slow because it has the authority of the sea behind it. Moreover, it has nothing to do with the joy and madness of escape from confinement, but daily performs a sober function of renewal. It is a deliberate, sightless creature, pushing before it sinuous fingers with which it gropes its way through the crushed jungles of matted weed. For the gulls, the redshanks, the stint, the herons, and the curlew, the important moments of the day are when the water first leaves the banks and a refreshed feeding-ground is once more laid bare. But to the yachtsman the vital time is when the sea advances, bringing its salt breath among the drowsier inland scents, raising the weed from the dead, and changing into sensitive buoyant things the smacks and yachts which have been stranded on their sides, heavy and immobile for hours. There are two yachtsmen at least who are almost ashamed to confess how childish in its reality is their pleasure in watching the return of the tide over the flats or up some shallow creek. They have not counted the number of times they have leaned over the side of a yacht, knowing she could not float for an hour or more, watching the tiny crabs scuttle into fresh territories as the oily flood bearing yellow flecks of tide-foam brims silently over one level on to the next; watching each weed being lifted and supported by the water until its whole length waves and bends in the tide like a poplar in a breeze; watching the angle at which the yacht has been lying correct itself until she sits upright in the mud; watching, perhaps, in the proper season, the swish and flutter of the water, and the little puffs of disturbed mud drifting away like smoke, as mullet thresh their way through the entrancing green submarine avenues. And then there is always the thrill of the moment when the rising water touches with life the dead hull of a yacht, and turns her into a creature of sensitiveness and grace swaying to the run of the tide. One moment she is as a rock against which you might push unavailingly with all your might; the next she has sidled off the ground, and will sheer this way and that in response to a finger laid upon the tiller.
- 73. As the tide rises towards its height you may see smacks—oyster dredgers, trawlers, shrimpers, and eel boats—filling the shining mouth of the estuary. The lighting of this part of the coast is like nothing else in England. A pearly radiance seems to strike upwards from the sea on to the underpart of the clouds, which borrows an abnormal glow. In these waters, when the sea is not grey it is generally shallow green, and sometimes, when there are thunder-clouds with sunshine, it becomes an astonishing jade. At sunset the vapours over the marshes burn like a furnace, and the cumulus clouds sometimes glow underneath with the dusky fire of a Red Underwing moth. When the water has left the flats the lighting does not change appreciably, because the gleaming mud, glossy and shining like the skin of the porpoises which sport along the channels, has the quality of water. The most characteristic effect is the mirage, which swallows up the meeting-point of sea and sky in a liquid glare, exalts the humblest smack with the freeboard and towering rigging of a barque, and separates the tops of trees from visible connection with the land, so that they appear to be growing out of air and water. Often one might fancy that the trees of the Blackwater and the Crouch, thus seen in the distance, were the palm-trees of some Polynesian island. On the marshes, or reclaimed lands, which are inside the sea-walls, and are intersected by tidal dykes called fleets, sea-fowl and woodland birds mingle: curlew with wood pigeons, plover with starlings, rooks and gulls, feeding harmoniously. Here and there the mast and brailed- up sail of a barge sticking out of grazing-land tell of a creek winding in from some hidden entrance, and remind you that in Essex agriculture and seamanship are on more intimate terms than are perhaps thought proper elsewhere. Outside the sea-walls are salt marshes (‘salts’ or ‘saltings’) which are covered only by the higher tides. In the early summer the thrift colours them with pink and white, and later a purple carpet is spread by the sea lavender. The juicy glasswort (called ‘samphire,’ though it is not the samphire of Dover Cliff in ‘Lear’) changes from a brilliant green to scarlet. Herons wade in the rivulets; the whistle of the redshanks, the mournful cry of the curlew, and the scream of the gulls which fringe
- 74. the edge of the water like the white crest of a breaking wave, sound from end to end of these marshes. In the winter you may hear the honking of Brent geese. But by far the most beautiful sight is hundreds of thousands of stint or dunlins on the wing together. These birds are also called ox-birds, and the fishermen call them simply ‘little birds.’ When they wheel, as at the word of command, the variations in their appearance are almost beyond belief; now they are wreathed smoke floating across the sky, and scarcely distinguishable from the long smudge that pours from the funnel of a steamer on the horizon; now the sun catches their white underparts, and they are a storm of driven snowflakes; now they present the razor edge of the wing, and then disappear in the glare as by magic; again they turn the broadest extent of their wings, and a solid and heavy mass blackens the sky. BEAUMONT QUAY
- 75. In May, when the sea-birds are hatching their young, the spring-tides are slack and do not cover the saltings. In a pretty figure of speech the fishermen call these tides the Bird Tides. The lives of the fishermen are ruled by the tides. For them the working hours of the clock have no significance. On the first of the ebb, be it night or day, their work begins, and it is on the flood that they return to their homes. They have no leisure or liking for the time-devouring practice of sailing over a foul tide. The tide in the affairs of these men is absolute. And although they do not confess in any recognizable phrase of lyrical sensation that the sea has cast a spell upon them, it is obvious that that is what has happened. On Sundays, when they are free from their labour, they will assemble on the hard—a firm strip of shingle laid upon the mud—and, with hands in pockets, gaze, through most of the hours of daylight, upon the sweeping tide and the minor movements of small boats and yachts with an air at once negligent and profound. The mightiness of the sea, like the mightiness of the mountain, draws mankind. Men have learned the secrets of these things in a way, and have turned them to their profit or amusement; but the mastery is superficial, and it is man who in these great presences is unconsciously and spiritually enslaved.
- 76. CHAPTER XIV ‘He was the mildest-mannered man That ever scuttled ship or cut a throat.’ A great merit of a barge as a house is that when she is ‘light,’ or almost ‘light,’ as the Ark Royal is, she can be sailed out of rough water on to a sand and left there, provided care be taken that she does not sit on her anchor. By the time there is only three feet of water the waves are very small, and thus, however strong the wind may be and however hard the sand, a barge will take the ground so gently that one can scarcely say when she touches. The explanation is simple enough, for, besides being flat-bottomed, a barge, owing to her length, strides many small waves at once. We put the plan into operation on our way to Newcliff. We were running up Swin, and with the dark the breeze piped up; so instead of sailing all night or anchoring in the Swin, where there would have been a disagreeable sea on the flood-tide, we put the Ark Royal on the sand between the Maplin Lighthouse and the Ridge Buoy, and there she sat as steady as a town hall. This is, of course, an easy way of going to the seaside, so to speak. You simply sail on to a nice clean sand and stay there till the wind moderates. Whenever the tide ebbs away, you can descend on to the sands by a ladder over the side, and pursue the usual seaside occupations of building docks and canals and forts and catching crabs. It was a memorable experience, this passage up the Thames estuary, house and furniture and family all moving together without any of the bother of packing up and catching trains, and counting heads and luggage at junctions. The children enjoyed every moment of it—the following sea and the dinghy plunging in our wake, the steamers bound out and in, the smacks lying to their nets with the gulls
- 77. wheeling round them waiting for their food, the tugs towing sailing ships, the topsail schooners, the buoys, the lightships. When we arrived at Newcliff we anchored off the town, intending to look for a good winter berth later in the year. After the quiet of Fleetwick, Newcliff struck us at once as over-full of noise and people. At all events, we had the satisfaction of knowing that we were not going to live on shore. The spot where we lay would have been well enough for the summer, though with a fresh breeze on shore it was impossible to take a boat safely alongside the stone wall. The boat, however, could be rowed up a creek half a mile away. Unfortunately, this meant the chance of being drenched with spray, and it was also a too uncertain way of catching trains and trams. Nine times out of ten we could row to the stone wall, and when the tide ebbed away and the Ark Royal lay high and dry (which, roughly, was for six out of every twelve hours) we could always walk ashore. The sand was hard under about an inch of fine silt. Here and there it was intersected by shallow gullies, but short sea-boots served our purpose of getting on shore dry. Of course, we always had to think ahead, for if one went ashore in the boat and took no sea-boots, it might be necessary on returning to walk to the Ark Royal; and if no one were on deck one might shout for sea- boots for a long time from the land before being heard. The most awkward time was when the flats were just covered with water, for then there was too much water round the Ark Royal for sea-boots and not enough to float a boat to the shore. Then one simply had to wait until it was possible to walk or row. Once we were caught in this way at one o’clock in the morning after going to a theatre in London. We waited a short time for the ebb, but were too sleepy to wait quite long enough. We put on our sea-boots; and then, slinging my evening shoes and the Mate’s round my neck, and cramming my opera-hat well on to my head, I gave the Mate my arm. The water itself was not too deep, but in the dark it was difficult to avoid the gullies, and the Mate nearly spoiled her new frock and my evening clothes by stumbling into a hole and clutching at me. This was the only occasion on which I should have been distressed if those who had disputed the advantages of living in a barge could have seen us. In anything like a gale of wind
- 78. there was a nasty, short, confused, broken sea, and then one had either to row up to the creek and be drenched or wait till the tide had ebbed. It was evident that lying off the town for the winter was out of the question. Soon we found a berth up the creek where yachts are laid up, and agreed to pay a pound for the use of it for a year. It was well sheltered, but as only a big tide would give us water into it we had to wait some days after we had found it. Meanwhile Sam Prawle, who had remained with us all this time, had to return home. The children had rallied him a good deal on his yarn about ’Ould Gladstone’ and on the ethics of salvage generally. Salvage was Sam Prawle’s favourite subject; and we could never make up our minds whether he was more given to boasting of what he had done or to regretting what he had not done. The evening before he went away he was evidently concerned lest he should leave us with an impression that salvage operations were not invariably honourable if not heroic affairs. He therefore related to us the following episode, and the reader must judge how far it helps Sam Prawle’s case: ‘In them days, afore it was so easy to git leave to launch the lifeboats as that is now, we allus used to keep a lugger for same as salvage work. The last wessel as ever I went off to on a salvage job my share come to thirteen pound and a bit extra for bein’ skipper, and if there hadn’t bin a North Sea pilot aboard that ship us chaps ’ud have had double. But then agin, if us hadn’t bin quick a makin’ our bargain us shouldn’t have had nawthen. ‘One night, after a dirty thick day blaowin’ the best part of a gale o’ wind sou-westerly, the wind flew out nor-west, as that often do, and that come clear and hard, so as when that come dawn you could see for miles. Well, away to the south’ard, about six mile, we seed a wessel on the Sizewell Bank; she was a layin’ with her head best in towards the land. There was a big sea runnin’, but there warn’t much trouble in launching the lugger with the wind that way, though we shipped a tidy sea afore we cast off the haulin’-aout warp.
- 79. ‘We’d close-reefed the two lugs afore we launched the bo’t, and it warn’t long afore the fifteen of us what owned the lugger was a racin’ off as hard as we dare. You see, we den’t want no one to git in ahead of we. Us dursn’t put her head straight for the ship, for the sea was all acrost with the shift o’ wind, and us had to keep bearin’ away and luffin’ up. You see, them seas was all untrue; they was heapin’ up, and breakin’ first one side, then t’other, same as in the race raound Orfordness. ’As we drawed near the wessel, that fared to we as haow she were to th’ south’ard of the high part of the sand, and that warn’t long afore we knaowed it, cos we got our landmarks what we fish by, for we most knaows that sand, same as you do the back o’ your hand, as the sayin’ is. We laowered our sails and unshipped the masts and raounded to under the wessel’s quarter—a barquentine, she were, of about nine hundred ton—and they thraowed us a line. All her sails was stowed ’cept the fore laower torpsail, which were blown to rags, and the sea was breakin’ over her port side pretty heavy. There warn’t no spars carried away, and there den’t fare to be no other damage, and if she was faithfully built she den’t ought to have come to a great deal o’ hurt so fur. ‘Then they thraowed us another line for me to come aboard by, and we hauled our ould bo’t up as close as we durst for the backwash. I jumped as she rose to a sea, but missed the mizzen riggin’ and fell agin the wessel’s side; them chaps hung on all right, and the next sea washed me on top o’ the rail afore they could haul in the slack. That fair knocked the wind aout o’ me, and I reckon I was lucky I den’t break nawthen. I scrambled up, and found the cap’n houldin’ on to the rail to steady himself agin the bumping o’ the wessel. ‘Well, she was paoundin’ fairly heavy, but not so bad as other wessels I’ve bin aboard. Still, that’s enough to scare the life aout of anyone what ain’t never bin ashore on a sandbank in a blaow, and most owners don’t give a cap’n a chance to do ut twice—nor pilots neither. I could see the cap’n fared wonnerful fidgety, for the wessel had been ashore for seven hours and more, so I starts to make a bargain with
- 80. him for four hundred pound to get his ship off, when up comes a North Sea pilot what was aboard. I was most took aback to see him there. ‘“What’s all this?” he says. ‘“Four hundred pound to get she off,” I says. ‘“Four hundred devils,” he says. ‘“No cure, no pay,” I says. ‘“No pay, you longshore shark!” he says. ‘Of course, he was a tryin’ to make out there warn’t no danger to the wessel and nawthen to make a fuss about. You see, he was afeared there might be questions asked about it, and he might get into trouble. Anyway, it don’t do a pilot no good to get a wessel ashore, even if that ain’t his fault which it warn’t this time, for the wessel was took aback by the shift o’ wind and got agraound afore they could do anything with her. ‘One thing I knaowed as soon as my foot touched them decks, and that was that she warn’t going to be long afore she come off. Sizewell Bank’s like many another raound here; that’s as hard as a road on the ebb and all alive on the flood, and them as knaows, same as we, can tell from the way a wessel bumps what she’s up to. I could feel she warn’t workin’ in the sand no more, but was beginning to fleet, and ’ud soon be paoundin’ heavier than ever, but ’ud be on the move each time a sea lifted she. Howsomdever, I kep’ my eyes on the cap’n, and I could see he was skeered about his wessel, and ’ud be suthen pleased to have she in deep water agin. ‘“Cap’n,” I says, “three hundred and fifty pounds. No cure, no pay.” “Too much,” says the cap’n, but I see he’d like to pay it. ‘“Too much?” says the pilot. “I should think it is! The tide’s a flowin’, and she’ll come off herself soon; besides, if she don’t we’ll have a dozen tugs and steamers by in two or three hours, and any of ’em glad to earn a fifty-pun’ note for a pluck off.”
- 81. ‘“That’ll be high water in two and a half hours, and you’ll be here another ebb if you ain’t careful,” I says to the cap’n, “and this sand’s as hard as a rock on the ebb. The pilot’ll tell you that if you don’t knaow that already for yourself.” ‘“There ain’t no call to pay all that money,” says the pilot. “She’ll come off right enough.” ‘“Well,” I says to the cap’n, “if I go off this ship I ain’t a comin’ aboard agin ’cept for much bigger money, and when she’s started her garboards and ’s making water you’ll be sorry you refused a fair offer!” ‘“I’ll give yer two hundred,” says the cap’n. ‘That fared to me best to take it, for she was bumpin’ heavier, and I laowed she’d begin to shift a bit soon. Then agin, the paounding was in our favour, for I see that skeered the cap’n wonnerful, so I starts a bluff on him. ‘“That ’on’t do, cap’n,” I says. “I’m off.” ‘I went to the lee side of the poop, where our ould bo’t was made fast, to have a look at my mates. The ould thing was tumblin’ abaout suthen, for there was a heavy backwash off the ship’s quarter. As she came up on a sea they caught sight o’ me and started pullin’ faces and shakin’ their heads, and next time I see them they was doin’ the same. I tumbled to it quick enough that they wanted to say suthen to me, and a course they couldn’t shaout it out, so I threw ’em the fall o’ the mizzen sheet, and me and one o’ the crew pulled ould Somers aboard. ‘“For ’eaven’s sake,” he says, close in my ear, “make a bargin quick! She’s a comin’ off by herself! We’ve got a lead on the graound, and she’s moved twenty foot already.” ‘I went back to the cap’n, and he was all on fidgetin’ worse’n ever, so I says, “Cap’n, my mates’ll be satisfied with three hundred paound.” ‘“Don’t you do no such thing,” says the pilot; “she’ll come off all right.” ‘“I’ll stick to my two hundred,” says the cap’n.
- 82. ‘I dursn’t wait, so I closed on it, and the mate writ aout two agreements, one for the cap’n and t’other for me. Our chaps soon got the kedge anchor and a hundred fathoms o’ warp into the lugger and laid that right aout astern, and I give the order for the lower main torpsail and upper fore torpsail to be set. ‘Then our chaps come aboard, and what with heavin’ her astern a bit every time she lifted to a sea and them two torpsails aback, she come off in half an hour. ‘Yes, yes; we got thirteen pound apiece, and if it hadn’t been for that pilot we’d a got double.’
- 83. CHAPTER XV ‘Mon Dieu, mon Dieu, la vie est là, Simple et tranquille; Cette paisible rumeur-là Vient de la ville.’ We engaged two men to help us up the creek, which is narrow and was full of small boats difficult for a large craft to avoid. Unluckily, there was no wind, and we had to punt. This made our difficulties greater, as the Ark Royal, unlike her trading sisters, could not cannon her way cheerfully up the creek lest her stanchions should be carried away or her cabin tops be damaged. The two men used the poles forward while I steered. A proud helmsman I was, knowing myself the owner and skipper of the largest yacht on the station, as we passed a quay thronged with longshoremen looking on. At that moment I had to put the wheel hard over, and as the barge’s stern swung towards the land her rudder touched the hawser of a smack moored at the shipyard. The pull of a ninety-ton vessel moving however slowly is enormous. The hawser tautened like a bar of iron; the Ark Royal’s rudder was banged amidships, wrenching the wheel from my hands; one of the spokes caught my belt, hoisted me off my feet, swung me right over the top of the wheel, and dropped me on the other side of the deck. The Mate and the children did not seem to understand that this accident to the Skipper reflected some ridicule on the whole ship’s company. They cackled with delight, and wanted me to do it again.
- 84. WALTON CREEK When we came abreast of our berth there was not enough water for us to go in, so we lay on a spit of sand and mud for that day. On the next tide, which was higher, we moved in stern first, leaving our anchor well out in the creek ready to haul us off in the spring. The ebbing tide left us in a shallow dock about three feet deep into which the Ark Royal just fitted, so that with a ladder on to the saltings we could easily get on and off the ship. From the road, seventy or eighty yards away, there was a path across the saltings right up to us, but as it was very muddy we bought forty or fifty bushels of
- 85. Welcome to our website – the perfect destination for book lovers and knowledge seekers. We believe that every book holds a new world, offering opportunities for learning, discovery, and personal growth. That’s why we are dedicated to bringing you a diverse collection of books, ranging from classic literature and specialized publications to self-development guides and children's books. More than just a book-buying platform, we strive to be a bridge connecting you with timeless cultural and intellectual values. With an elegant, user-friendly interface and a smart search system, you can quickly find the books that best suit your interests. Additionally, our special promotions and home delivery services help you save time and fully enjoy the joy of reading. Join us on a journey of knowledge exploration, passion nurturing, and personal growth every day! ebookbell.com