AN ADAPTIVE APPROACH FOR DYNAMIC RECOVERY DECISIONS IN WEB SERVICE COMPOSITION USING SPACE BASED QOS FACTOR

International Journal on Web Service Computing (IJWSC), Vol.6, No.2/3, September 2015
DOI : 10.5121/ijwsc.2015.6302 11
AN ADAPTIVE APPROACH FOR DYNAMIC RECOVERY
DECISIONS IN WEB SERVICE COMPOSITION USING
SPACE BASED QOS FACTOR
Dr. Ravi Shankar Pandey, Richa Pathak
Department of Computer Engineering, BIT Mesra , Ranchi, India
ABSTRACT
Service Oriented Architecture facilitates automatic execution and composition of web services in
distributed environment. This service composition in the heterogeneous environment may suffer from
various kinds of service failures. These failures interrupt the execution of composite web services and
lead towards complete system failure. The dynamic recovery decisions of the failed services are
dependent on non-functional attributes of the services. In the recent years, various methodologies
have been presented to provide recovery decisions based on time related QoS (Quality of Service)
factors. These QoS attributes can be categorized further. Our paper categorized these attributes as
space and time. In this paper, we have proposed an affinity model to quantify the location affinity for
composition of web services. Furthermore, we have also suggested a replication mechanism and
algorithm for taking recovery decisions based on time and space based QoS parameters and usage
pattern of the services by the user.
KEYWORDS
Failure recovery, QoS Factors, Location Affinity, Recovery Decisions
1. INTRODUCTION
Enterprises are using information technology to automate their business activities within the
organization as well as with other organizations. They are using different software and hardware
platforms for automating these activities. In general these business activities consist of more than
one smaller sub activities. These activities may reside at same geographical location or at different
geographical location. Internet provides an infrastructure for integrating such activities. A new
paradigm is developed to facilitate these integrations known as Service Oriented Architecture
(SOA). In this architecture software’s are viewed as services. This architecture contains
description of the service using WSDL (Web Service Description Language), communication
protocol SOAP (Simple Object Access Protocol) for providing interaction among the services and
UDDI (Universal Description, Discovery and Integration) storage mechanism for these services.
All they are based on open source mechanism provided by XML (Extensible Markup Language).
These services are loosely coupled in nature and may be integrated at run time dynamically. This
architecture facilitates three types of entities; these are service requester, service provider and
service broker. The main role of the service providers are to host their services in storage
area(UDDI) and requester searches the desired service through broker in automated environment.

12
Different service providers may publish the same services in the UDDI registry. So it become
strenuous to select a single service from the same service domain based on their functional
features. In these view non-functional properties of the services play an important role in service
selection. These non functional properties are known as QoS attributes. The service providers
consider these parameters as QoS attribute for their published services while service requester uses
these parameters as non-functional requirement for their desired services. These parameters are
availability, execution time, cost, reliability, latency. All these parameters are primarily based on
time factor. .[24] have proposed QoS attributes which are based o the usability of the services in
their composition other than time related QoS (dominant role, dominant operations).Web services
are hosted at different geographical locations and are used by people situated at different
geographical regions. In this scenario services may be preferred by their geographical location or
they may be frequently used for a particular region. Thus we have concluded that for service
selection location is also a non-functional attribute of a service. We are categorizing all the non-
functional attributes or features of web service in two categories. One is the QoS based on time
factor (reliability, cost, latency, execution time) and other is the QoS based on location that is
location affinity of the service. We have introduced a term “location affinity” with respect to the
web services as a space based quality of service factor. We are describing the location affinity as
the degree or extent to which user likes and utilize the service of a particular geographical location.
When a service execution takes place in atomic manner it is less susceptible to failure. Service
composition is prone to failure due to lack of availability, semantic mismatching and other
functional attributes. Several research efforts have been made to model non-functional parameters
and provide a recovery mechanism in case of the failure occurred during compositions.
In this paper, we have proposed a methodology for taking the recovery decisions to handle the
failure in service composition based on time as well as location based QoS attributes together. We
are also proposing finite state machine model for the web service composition. This model
includes their QoS parameters, input parameters, set of operations, transition function for transition
from one state to another and evaluating impact on Overall QoS factors of our model. We are also
suggesting a methodology to take recovery decisions dynamically.
This paper is organized as follows. The section 2, introduces the related work on service
composition, service failure, QoS evaluation using finite state machines. In section 3, we
represented the location affinity based meta-model and finite state machine model for service
composition using QoS attributes. In the last section, we presented a replication strategy and
algorithm to take recovery decision in the presence of fault.
2. RELATED WORK
Now a day’s web based services are being capturing the IT market completely. With the growth of
use of Internet user’s requirement is also growing explosively and single service fails to meet this
requirement. Thus service composition came in to the existence. Various methods for service
composition have been presented whether it is dynamic or static. Sunil R Dhore et al [1] have
discussed the semantic composer that uses enhanced ant colony optimization mechanism to
provide service composition in efficient manner by finding optimal composition length in each set
of candidate web service proposed for service composition. Zhou Xiangbing et al [2] have
presented a web service modeling ontology for service composition using genetic algorithm. In
this paper they have considered four non functional attributes, these are: information exchange,

13
time, matching reasoning and cost. Finite state machine based composition solutions have also
been proposed by different authors.Olga et al [4] have estimated the quality of service and quality
of experience of web service using the finite state machine concept. Jun Sun et al [25] have
suggested the finite state machine synthesis for software systems. Number of methods has been
proposed to monitor and identify faults in service based systems. Jocelyn Simmonds et al [9] have
presented an approach to monitor and recover web service based applications using three stage
phenomenons. In first stage of preprocessing BPEL code is converted in to labeled transition
systems and properties supplied by the user are converted in to monitor code .In monitor phase
application is checked for an error and In recovery phase compensation plan are ranked and
applied. Compensation allows a web service to go-back to the previous stage if an error occurred.
Thirumaran.M et al [6] have proposed a QoS based run time exception handler. This handler first
takes exception information as an input and on the basis of that calculates the time and space
complexity. Decidability evaluator than categorize the problem in to, NP-Complete solvable
problem set, NP-Hard may be solvable problem set. They have also used turning machine to find
the solution of the problem in a specified time using audit log that contains some predefined
solutions for particular problem. Services composed at run time are more prone to failure. These
failures could occur due to various reasons .Hadi Saboohi et al [11] have figured out reasons for
the failures. This paper identifies the functional as well as non functional causes of service
composition failure. Functional causes includes service unavailability, service malfunctioning and
non functional causes includes unexpected data, network delay, response time-out. Abdelkarem
Erradi et al [17] have proposed various recovery policies (retry, substitution, parallel execution,
dynamic update of service composition). They have also proposed MASC (Manageable &
Adaptive Service Composition) model for monitoring failure and providing recovery from failure.
H. Elfawal Mansour et al [14] have proposed a model that calculates the reliability at run time and
uses roll back mechanism to recover services from failure. They have used a broker that decides
that the result computed by a particular web service is relevant or not. If it is relevant process of
computation continues otherwise the service is roll backed and another service having same
functionality is been called to complete the execution of the complete web service based system.
Suchi Gupta et al [8] have presented subset replacement mechanism to handle service failure.
According to this paper failed set of services are first identified by the middle agent and then
replaced by another set of services providing the similar functionality. Keting Yin [15] have also
used replacement mechanism to recover service from failure. They have also presented a ranking
mechanism to select the service that could replace the failed service based on non functional QoS
parameters. Guisheng Fan et al[16] has categorized the service failure as failure of available
service, failure of component service, Failure of operating environment and model the composite
web services using petri-nets. Cao et al [18] have presented in which firstly, the service execution
graph (SEG) is introduced and a service execution solution selection algorithm is proposed to
choose one from the solution set of TSSA which has the highest success rate of recovery. Then,
the concepts of execution backup path and switch cost are introduced and a search algorithm is
presented to search for the optimal backup path when current service failure occurs.Rafael
Angarita et al [10] have categorized the recovery mechanism in to two parts based on the ACID
properties of a transaction. These categories are forward and backward recovery mechanism. In
forward recovery mechanism replacement, compensation and replication strategies are included. In
backward recovery mechanism roll-back strategy is included. All the recovery mechanism has
been tested based on the execution cost and execution time in different failure conditions and
dynamic recovery decisions could be taken on the basis of experimental result. Replication of the
web service is the preferable solution of handling failure of the services in distributed
environment. Marwa F. Mohamed et al [21] have provided a framework for the dynamic

14
replication of the web services .This framework automatically replicates a service based on it’s
consumption and reduces the service response time as well. They have taken sensor based
mechanism to identify the service failure and then take replication decision for the failed web
service. Mario Bravetti et al [22] computed the system performance after replicating the web
service using SOCK calculus. An Liu et al [26] have proposed a framework named as “FACTS” to
provide highly reliable and fault tolerant service composition. They have also used BPEL to
elaborate their proposed architecture.Johannes Behl et al [27] have presented architecture to
replicate the BPEL engine as well as web service using proxy servers (input and output proxy
servers) to enhance the availability and reliability of the services in cloud computing platform.
Ivona Brandic et al [23] location affinity QoS with respect to the Grid Computing environment. It
is basically used for providing the security to the users of the grid and giving the grid resources of
particular domain (organization, home) based on the user’s willingness. R.S.Pandey et al [24] has
proposed a methodology to estimate the minimum and maximum value of the Quality of service
attributes (reliability, availability, latency, security). We have analyzed that all the above
researches basically emphasis upon the time based Quality of Service factors for composing the
service and also for recovery decision. In [10] recovery decisions were taken dynamically on the
basis of the execution time and execution cost. None of the author considered the geographical
location factor to take recovery decisions at run time. In our paper we are selecting recovery
strategy among all the available strategies (replacement, roll-back, replacement) based on
geographical location .In this paper we are also advocating the need of the replication mechanism
and suggested a methodology to decide which service from the same location affinity and similar
set of functionality should be replicate in distributed environment to avoid failure. In our proposed
work we are including location factor for the composition of services based on finite state
machine. Dynamic recovery concept has also been extended with location necessity.
3. MODELING LOCATION AFFINITY
In [23] location affinity is modeled to benefit the grid computing users. The authors of this paper
mainly concerned with security and legal issues, while using grid technology. They have also
discussed that grid users hesitate to use this technology everywhere rather they wanted to use this
technology within a specific domain or range. They have modeled location affinity for grid user
with in a specific domain of their choice for the security purpose. QoWL (QoS-aware Grid Flow
Language) that is a XML based language is developed and used for modeling the location affinity.
In [23] location affinity is modeled for the security purposes but in our paper we are modeling
location affinity with a different perspective. We are considering affinity as the user’s preference
to use a particular service that belongs to a certain geographical boundary. For example a user
weekly visits to the Mumbai city and requires hotel service of Mumbai city frequently. Thus user’s
location affinity is Mumbai city. In this model, we have considered that universe is divided into the
continents, continents are divided into countries, countries are divided into the states, states are
divided into the district and districts are divided into cities and villages respectively. This model
uses various attributes to distinguish among countries, states, cities and villages. As per our model
city is specialized geographical area of the district and has several villages. Similarly, state is an
aggregation of all the districts which belongs to the same state. In this model, we have not
distinguished different types of states for simplicity. We have not further detailed city and village
to make the model simple.Fig1 is representing the meta-model of location affinity. We are going to
use this model in web service composition. In web service composition several operators are used
(sequential, parallel, loop and switch). Service may be composed one after other which is called as

15
sequential execution of the services or two services may execute simultaneously in parallel mode.
One service may execute more than one time or under any defined condition. The proposed model
has some pre-assumptions for service composition based on our proposed geographical criteria’s.
If two services are executing in the same state than their affinity is considered to be the city or
district affinity. Similarly, if these services are composing from different states than their affinity is
considered as state affinity. Consider that service1’s location affinity is Allahabad city of Uttar
Pradesh state and services2’s location affinity is the Banaras city of the state Uttar Pradesh. Then,
while composing their affinity is city since lying in the same state or region of our model.
Similarly, we have considered that service1’s location affinity is Allahabad city of Uttar Pradesh
state and Service 2’s location affinity is Mumbai city of the State Mumbai then compositely their
affinity would be state since lying in different boundaries of our model.
Figure 1. Modeling Location Affinity
4. FINITE STATE MACHINE FOR REPRESENTING WEB SERVICES
Every web service consists of several input and output operations. A service changes its state due
to the invocation of various available operations. These invocations may affect the parameters of
quality of service attributes. For example: if the invocation is successful then reliability of web
service will be higher. The other attributes also may change due to the state change like cost. For
example: In ticket Reservation system if we invoke bookTicket operation, in this scenario the cost
of the web service may change. In the case of single web service execution, the space related QoS
attributes like affinity will not change. Several research efforts have been made to incorporate
time-based QoS attributes to model web service using finite state machine [4], [24]. We are
arguing that the space related QoS attribute will play a dominant role in the case of web service
composition. We have modeled a web service as finite state machine that includes set of input
operations, initial state, final state, transition function and quality of service parameters. Our

16
model is based on time as well as space related QoS attributes, so it is a vector which has two
different types of values one is based on geographical location and other is based on time. These
component services have been modeled using finite state machine concept. A component web
service is a finite state machine having touple (I, O, F, S0 , SF, δ, Q).Where
I is representing the set of input.
O is representing the set operations
δ is the transition function that is represented as :
δ: I × O × S× Q S
SF is the final state of the finite state machine.
F is the subset of S and represents the set of final states.
Q is the quality of service factor which is categorized on the basis of the time and
space dimensions i.e. Qs ɛ Q and QT ɛ Q as well as
Q=QT × QS
QT=QE × QR × QA
Where QE is the execution time, QR is for reliability and QA for availability.
Affinity is the quality of service attribute based on location of the web service. It is computed by
the user’s usability of web service based on user’s location preferences.
5. MODELING COMPOSITION OF WEB SERVICES BASED ON LOCATION
AFFINITY
In [4] a finite state machine based model is proposed to evaluate QoS parameters of service
composition. They have also elaborated how machine learning algorithms can be used for
evaluating QoS of service composition. [25] have proposed a model for synthesizing finite state
machine. We are synthesizing finite machine based on time as well as space based QoS
parameters.
A service could individually serve the user or it could be the composition of two or more services.
Thus, overall system output is dependent on the successful execution of the each component web
service existing in composition plan.
To give rules for composition of web services we are considering two state machines M1 and M2
i.e. M1= (I1, O1, F1, S01, SF1, δ1, Q1) and M2 = (I2, O2, F2, S02 , SF2, δ2, Q2). Resultant machine can
be expressed as, M = (I, O, F, S01 ,SF1, δ, Q) .Where
I=I1 U I2
O=O1 U O2
F = F1 × F2
δ: I1 × I2 ×O1 ×O2 × Q × S1× S2 F1 × F2
Q Q1 composition Q2
In composition model services may execute in sequential manner, parallel manner, in the form of
loop and choice [24], [29].

17
Thus quality of service attributes composition depends upon the composition operators. We are
using the composition rules for all the composition operators as given in [24], [29]. We are
summarizing these rules as given below with respect to the two services S1 and S2.
We have assumed that (r1, A1, E1) is the reliability, availability and execution time of the service S1
and (r2, A2, E2) is the reliability, availability and execution time of the service S2 respectively.
Consider that services S1 and S2 are executing in sequential manner than according to the rules
given in [24],[29] resultant reliability is the product of reliabilities of both the services, resultant
availability is the product of availabilities of both the services and resultant execution time is the
summation of the execution times of both the services respectively. Our affinity rule suggests that
if two services have same level of affinity than the resultant affinity will be one level higher than
the previous level affinity.
QT=QT1 × QT2= (r1, A1, E1), (r2, A2, E2) =(r, A, E)
r= r1 × r2
A=A1 × A2
E=E1+E2
Consider that services S1 and S2 are executing in parallel manner than according to the rules
[24,29] resultant reliability is the minimum of the reliabilities of both the services, resultant
availability is the minimum of the availabilities of both the services and resultant execution time is
the maximum of the execution times of both the services respectively. If the parallel execution of
web services is taking place during composition than the resultant affinity level is considered to be
at same affinity level.
QT=QT1 || QT2 = (r1,A1,E1)||( (r2,A2,E2)=(r,A,E)
r = r1|| r2=minimum (r1, r2)
A=A1||A2=minimum (A1, A2)
E=maximum (E1, E2)
Consider that services S1 and S2 are executing in loop than according to the rules given in [24,29]
resultant reliability is the minimum of the reliabilities of both the services, resultant availability is
the minimum of availabilities of both the services and resultant execution time is the multiple of
the number of times loop is executing a service respectively. In loop based execution of web
services the resultant affinity level is considered to be as before the execution of the service.
QT=QT1 loopQT2 = (r1,A1,E1)loop( (r2,A2,E2)=(r,A,E)
r= r1= r2
A= A1=A2
E=n × E
Where n is the number of times loop is executing.
Consider that services S1 and S2 are executing based on choice than according to the rules [24],[29]
resultant reliability is the minimum of the reliabilities of both the services, resultant availability is
the minimum of availabilities of both the services and resultant execution time is the maximum of

18
the execution time of both the services respectively. In choice operator the affinity is the affinity of
the web service that is selected for the execution among all available choices.
QT=QT1 choice QT2 = (r1,A1,E1)or( (r2,A2,E2)=(r,A,E)
r =min (r1, r2)
A=min (A1, A2)
E=max (E1, E2)
Example 1.
Consider the Figure 2, which represents the composite execution of the purchase order service.
This service uses four services for completing the execution plan. These services are defined as:
Buyer:
This service is agent service that initiates the composition plan and invokes seller service after
giving the request of particular product to purchase.
Seller:
This service receives the item request from Buyer service and process the request by initiating the
execution of InventoryCheck and CreditCheck services. After receiving the response returned by
both the services Seller acknowledge the Buyer about the status of his request of product purchase.
InventoryCheck:
This service is responsible for checking the quantity of the product ordered by the Buyer. This
service ensures that the quantity of item requested by the Buyer is available in the stock, if not it
sends a negative acknowledgement to the Seller service to maintain the consistency.
CreditCheck:
This service checks the Buyer’s balance and acknowledge the seller, whether Buyer is having
sufficient amount to purchase the item or not.

19
Figure 2. Composite Purchase Order Service
5.1. Aaa Finite State Machine for Buyer Service
Finite state machine takes the name of the product as an input and request product order. In this
description of the finite state machine pro_name is representing name of the product ordered by
Buyer. Here reqOrder () operation is used to send the product order request to the Seller. Few
variables are also used to represent the communication process among services. In this finite state
machine representation poreqstate variable is used to store the status of the request whether it is
sent from the Buyer’s end to the Seller’s end or not and r1,a1,e1,la1 variables are used to store
reliability, availability, execution time and location affinity of the Buyer service.
• I= {pro_name}
• O= {reqOrder ()}
• δ = {pro_name} × {reqOrder()} × S11{poreqstate, r1, a1,e1, la1} S12{poreqstate =”Sent”,
r1=.7, a1=.8 ,e1=8, la1= “Lucknow”}
• QT= {.7, .8, 8}
• QS= {“Lucknow”}
5.2. Bbb Finite State Machine for Seller Service
Finite state machine of Seller service takes the name of the product and quantity of the product as
an input. In this description of the finite state machine pro_name and quantity are representing
name of the product and quantity of the product ordered by Buyer. processOrder () operation is
used for further processing in the composite system. Variables poreqstate, ccrequeststate,
ivreqstate,poack are used to store the status of the product order request state, credit check request
state, inventory check request state In this scenario.r2,A2,E2,la2 variables are used to store
reliability, availability, execution time and location affinity of the Seller service.
• I= {pro_name, quantity}
• O= {processOrder ()}

20
• δ={pro_name,quantity}×{processOrder()}×S21{poeqstate,ccrequeststate,ivreqstate,poack,
r2,A2,E2,la2} S22{poreqstate=”received”,ccreqstate=”sent”,ivreqstate=”sent”,poack=”sen
t”, r2=.60,A2=.55,E2=5,la2= “Allahabad” }
• QT= {.60, .55, 5}
• QS= {“Allahabad”}
5.3. Ccc Finite State Machine for InventoryCheck Service
Finite state machine takes the quantity of the product as an input. checkInvAvailability ()
operation is used to check whether the required quantity of the product is available in inventory
stock or not. In this context ivreqstate, invavailablestate, invack, r3, A3, E3, la3 notations are used
to describe inventory request state, inventory availability state, inventory acknowledgement state,
reliability, availability, execution time, location affinity of InventoryCheck service.
• I= {quantity}
• O={checkInvAvailability()}
• δ = {quantity}× {checkInvAvailability()} × S31{ ivreqstate
,invavailablestate,invack,r3,A3,E3,la3}
S32{ivreqstate=”received”,inavailablestate=”true”,invack=”sent”, r3=.7,.A3=.6, E3=7,
la3= “Mumbai” }
• QT= {.7, .6, 7}
• QS= {“Mumbai”}
5.4. Ddd Finite State Machine for CreditCheck Service
Finite state machine takes the account detail of the buyer as an input. In this finite state machine
checkAvailability () operation is used to check the required balance for purchasing the product. In
this scenario ccreqstate, ccavailablestate, ccack, r4, A4, E4, la4 notations are used to describe credit
check request state, credit availability state, credit acknowledgement state, reliability, availability,
execution time, location affinity of Credit Check service.
• I= {account_detail}
• O={creditAvailability ()}
• δ = {account_detail}× {checkAvailability() }× S41{ ccreqstate,ccavailablestate,
ccAck,r4,A4,E4,la4}
S42{ccreqstate=”recieved”,ccavailablestate=”true”,ccack=”sent”,r4= .9,A4= .8
,E4=11,la4=” Banaras” }
• QT= {.9,.8,11}
• QS= {“Banaras”}
5.5. Eee Service Composition
These Services have been composed to complete the system execution. In this example Buyer
service is executing first in sequential manner with Seller service. Seller service than invoke the
InventoryCheck and CreditCheck services in parallel to give desired output.

21
Buyer.Seller.InventoryCheck||CreditCheck
The resultant Quality of service and finite state machine of composite service purchase order can
be represented as given below. Where IB.S , QTB.QTS, QSB.QSS are representing input , time based
QoS, space based QoS of Buyer andSeller together.
• I B.S= {IB U IS} = {pro_name, quantity}
• OB .OS= {OBUOS} = {reqOrder (), processOrder ()}
• δB.δS=IB×IS×OB×OS× S11×S21 S12×S22= {pro_name} × {quantity} × {reqOrder()} × {
processOrder()} ×S11{poreqstate , r1 , A1 , E1 , la1} × S 21{poreqstate , ccrequeststate ,
ivreqstate , poack , r2 , A2 , E2 , la2} S12{poreqstate=”Sent”, r1=.7, A1=.8, E1=8,
la1=”Lucknow”} × S22{poreqstate=”received”, ccreqstate=”sent”, ivreqstate=”sent”,
poack=”sent”, r2=.6, A2=.55, E2=5, la2=” Allahabad“}
• QTB.QTS= {.42, .44, 13}
• QSB.QSS = {“STATE AFFINITY”}
Since Inventory and Credit Check services are executing in parallel manner
WI||WC
Finite Sate machine for composition have been represented as:
• I CI.CC= {ICI U ICC} = {quantity, account_detail}
• O CI.CC= {OCI U OCC } = {checkInvAvailability() , checkAvailability ()}
• δCI .δCC=ICI × ICC × OCI×OCC× S31×S41 S32×S42={Quantity} × { Account_detail } × {
checkInvAvailability()} ×{checkAvailability ()} × S31 { ivreqstate , invavailablestate,
invack , r3, A3, E3, la3} × S41{ ccreqstate , ccavailablestate, ccack , r4 , A4 , E4 ,
la4} S32{ivreqstate=”received”, inavailablestate=”true”, invack=”sent”, r3=.7, A3= .6,
E3=7 , la3= “Mumbai”} × S42{ccreqstate=”received”, ccavailablestate=”true”,
ccack=”sent” , r4= .9 , A4= .8 , E4=11 , la4=”Banaras” }
• QTS.QTC={ .7, .6, 11 }
• QSI.QSC= {“STATE AFFINITY”}
In this finite state model ICI.CC , O CI.CC , δCI .δCC , QTI.QTC , QSI.QSC are representing the
composite input, output, transition function ,time based QoS and space based QoS of the
composite Inventory and Credit check services. Complete composite finite state machine for
purchase order service is given below:
• I= {pro_name, quantity, account_detail}
• O= {reqOrder (), processOrder (), checkInvAvailability(),checkAvailability ()}
• δ = {pro_name, quantity, account_detail}×{ reqOrder (), processOrder (),
checkInvAvailability(), checkAvailability()} ×S1f{poreqstate, r1, A1, E1, la1,
ccrequeststate, ivreqstate, poack, r2, A2, E2, la2, ivreqstate , invavailablestate, ccreqstate,
ccavailablestate, ccack, r4, A4, E4, la4, invack, r3, A3, E3, la3} S2f{ poreqstate=”Sent”,
r1=.7,A1=.8,E1=8, la1=”Lucknow”, poreqstate=”received”, ccreqstate=”sent”,
ivreqstate=”sent”, poack=”sent”, r2=.6,A2=.55,E2=5,la2= “Allahabad”, ivreqstate
=”received”, inavailablestate=”true”, invack=”sent”,r3=.7, A3=.6, E3=7, la3= “Mumbai”,
ccreqstate=”received” , ccavailablestate=”true” , ccack=”sent”,r4= .9,A 4= .8, E4=11,
la4=”Banaras”}
• QTB.QTS= {.29, .26, 24}

22
• QSB.QSS = {“COUNTRY AFFINITY”}
Thus
of the service having similar functionality and similar affinity than we send the replica
copy for the location affinity of the complete composite service execution is considered to be
“Country” and in case of the failure service providing the same functionality from the country
affinity is suggested to replace with failed service.
6. REPLICATION POLICY
In the view of advancement in distributed computing, need of replication is also growing rapidly
Service replication is also the most popular remedy to avoid failures during service execution. In
the scenario where a system contains only three and four services full replication gives 90% fault
tolerance. In case of the system having thousands of services, replication is not a feasible choice
for us due to storage and processing overhead.
So in distributed computing full replication is a very costly affair. It generally added measurable
amount of cost in total composition cost. We are providing a methodology according to which
only one service from the same functionality domain as well as location affinity is replicated to
reduce the storage overhead of the system. Replication of a service is done on the basis of the
threshold value. This threshold value is considered as the popularity factor of the service from the
same location affinity. This threshold value is calculated on the basis of the service usage pattern
of the user.
Definition 1. Popularity index is the index according to which replica of a service is being
created. This popularity
is dependent upon the average usage amount of the service by the users in the System. Notation
for popularity factor is P and U for average service usage amount.
P ∝U (1)
P= RU (2)
Where R is the constant and depends upon the ratio of the number of users of service and total
users of the system. [28] have also modeled replication manager concept to assign jobs to the web
services based on round robin algorithm for proper utilization of resources. In this research we are
proposing replication manger concept to calculate the popularity index and creating replica of
most popular service.
Consider a system that contains the five users of the seller services from the same location affinity
that is Mumbai (Maharashtra).There are five seller services from the same location affinity that are
providing similar set of functions to the user. These services are Seller1, Seller2, Seller3, Seller4,
and Seller5 respectively. In the system there are 500 users. Replicating all the services will increase
the performance as well as storage overhead .Solution of this problem is the calculation of
popularity factor. Service having highest population factor is being replicated to avoid the fault in
the system.

23
Table 1. Service Usage Frequency
Popularity factor is calculated for all the service given in table 1.
P SELLER1 = (5/500)×(750/5)=1.5
PSELLER2 = (5/500) ×(380/5)=.76
PSELLER3= (5/500) × (850/5) =1.7
PSELLER4= (5/500) × (500/5) =1
PSELLER5= (5/500) × (400/5) =.80
From the calculations given above Seller3 is identified as the highest popularity service. This
service is replicated in the distributed environment. Periodic update in the service usage pattern is
performed by replication manager to find out the highest popular service in the service domain and
particular location affinity.
6.1. Aaa PROPOSED ALGORITHM
Taking in to consideration the concept given in the replication policy we are assuming that we
have already a replica of the service providing similar services from the same affinity as a failure
handling mechanism. Our suggested algorithm is taking three types of the recovery decisions
.Firstly it checks the replica copy of the failed service if we found the replica execution. If it found
replica copy as unavailable as it takes the replacement approach as a remedy of failure. Lastly if
both the foresaid operations are unsuccessful it just roll-back the execution of the service.In this
scenario r1,a1,E1, is representing the reliability, availability ,execution time of the failed service
and r2,a2,E2 reliability, availability ,execution time of the service to be replaced.
Service
/User
Seller1
(MUM)
Seller 2
(MUM)
Seller3
(MUM)
Seller4
(MUM)
Seller5
(MUM)
User1
(MUM)
100 100 300 100 100
User2
(MUM)
200 50 100 100 100
User3
(MUM)
300 30 300 100 50
User4
(MUM)
50 150 100 100 50
User5
(MUM)
100 50 50 100 100

24
Algorithm 1. Service Recovery Based on time as well as space QoS
Input: Failed Service
Output:Selected Recovery Mechanism
1. Check for the availability of the replica of the service.
2. If (Replica Available) then
3. Provide replica of the most popular service providing similer functionality from the same
location affinity to the user.
4. Else
5. If (r1<r2 && a1<a2 && E1 >E2&& user’s
location affinity==location Affinity of Service)
Then
6. replace the service
7. Else
8. Rollback the failed service
7. CONCLUSION
In this paper we have modeled geographical locations for computing location affinity of the
services. We have also integrated time based QoS attributes and space based QoS attributes to
model failure. A service failure recovery policy is also introduced that takes recovery decisions
dynamically based on our suggested QoS attributes. Replication is the basic demand of
distributed systems but it is a costlier effort to implement. Thus we have initially included a
popularity value to decide the service that should be replicated. We also suggested the algorithm
that takes recovery decisions based on time as well as space based QoS parameters.
REFERENCES
[1] Sunil R Dhore, M U Kharat,”QoS Based Web Service Composition using Ant Colony Optimization:
Mobile Agent Approach”, International Journal of Advance Research in Computer and
Communication Engineering, vol. 1, issue 7, pp. 519-527, September 2012.
[2] Zhou Xiangbing, Ma Hongjiang, Miao Fang “An Optimal Approach to the QoS-Based WSMO Web
Service Composition Using Genetic Algorithm”, Workshops Lecture volume 7759, pp 127-139,
Springer, 2013.
[3] Quan Z. Sheng a, xiaoqiang Qiao b,, Athanasios V. Vasilakos c, Claudia Szabo a,Scott Bourne a,
xiaofei xu d ,“Web services composition: A decade’s overview”, Information Sciences 280 ,pp. 218–
238,Elsevier,2014.
[4] Olga kondratyeva,Natalia Kushik ,Ana Cavalli,Nina Yevtushenko,” Evaluating Web Service Quality
using Finite State Models”, Proc. 13th International Conference on Quality Software, pp.95 – 102,
IEEE, 2013.
[5] Kevin Wiesner, Roman Vacul´ın, Martin Kollingbaum, and Katia Sycara, ”Recovery Mechanisms for
Semantic Web Services”, Distributed Applications and Interoperable Systems ,Lecture Notes in
Computer Science, volume 5053, pp 100-105,Springer,2008.
[6] Thirumaran.Ma,Dhavachelvan.Pb,Shanmugapriya.Rc, Kiran Kumar Reddyd,” A Novel Approach for
Web Service Run Time Exception Handling”, In the proc. of 2nd International Conference on
Communication, Computing & Security, volume 6, pp.145–152, Elsevier,2012.

25
[7] Jocelyn Simmonds, Shoham Ben-David, Marsha Chechik,” Guided Recovery for Web Service
Applications”, Proc. eighteenth ACM SIGSOFT international symposium on foundations of software
engineering, pp. 247-256, 2010.
[8] Suchi Gupta, Praveen Bhanodia,”A Fault Tolerant Mechanism for composition of Web Services using
Subset Replacement”, International Journal of Advance Research in Computer and Communication
Engineering, vol. 2, issue 8, pp. 3080-3085, August 2013.
[9] Jocelyn Simmonds, Shoham Ben-David, Marsha Chechik, “Monitoring and Recovery of Web Service
Applications”, volume 95,issue 3,pp. 223-267,computing ,2012.
[10] Rafael Angarita, Yudith Cardinale, Marta Rukoz,”Reliable Composite Web Services Execution:
Towards a Dynamic Recovery Decision”, Electronic Notes in Theoretical Computer Science, vol.302,
pp.5-28, Elsevier, 2014.
[11] Hadi Saboohi, Sameem Abdul Kareem,” Requirements of a Recovery Solution for Failure of
Composite Web Services”, International Journal of Web & Semantic Technology, vol.3, issue 4, pp.
3-19, October 2012.
[12] Zho Wu,Naixue xiong,Wenlin Han,Yan N. Huang,Chun Y. Hu,Qiong Gu, Bo Hang, ”A Fault-
Tolerant Method for Enhancing Reliability of Service Composition Application in WSNs Based on
BPEL”, International Journal of Distributed Sensor Networks ,volume 2013,arcticle id 493678,II
page.
[13] Hossein Rahmani, Hassanabolhassani, ”Composite Web Service Failure Recovery Considering User
Non-Functional Preferences “, Proc. of 4th International Conference on Next Generation Web
Services Practices, pp. 39 – 45, IEEE, 2008.
[14] H. Elfawal Mansour and T. Dillon, Fellow, IEEE,” Dependability and Rollback Recovery for
Composite Web Services”, IEEE TRANSACTIONS ON SERVICE COMPUTING, vol. 4, pp. 328 –
339, October-December 2011.
[15] Keting Yin , Bo Zhou , Shuai Zhang , Bin Xu “QoS-Aware Services Replacement of Web Service
Composition”, Proc. of Internal Conference on Information Technology and Computer Science,
pp.271 – 274, IEEE,2009.
[16] Guisheng Fan, Huiqun Yu, Liqiong Chen, Chunhua Gu1,” An Approach to Handling Failure
Recovery in Service Composition and Its Analysis”, Proc. of Fifth IEEE International Conference on
Theoretical Aspects of Software Engineering, pp. 153 – 160,2011.
[17] Abdelkarim Erradi, Piyush Maheshwari2, Vladimir Tosic,”Recovery Policies for Enhancing Web
Services Reliability”, Proc. IEEE International Conference on Web Services, pp. 189 – 196, 2006.
[18] Cao Jiuxin, Zhou Tao, Zhu Gongrui, Liu Bo, Luo Junzhou, “Execution Recovery in Transactional
Composite Service”, Proc. 20th International Conference on Web Services, pp.276-283, IEEE,2013.
[19] Joyce El Haddad, Maude Manouvrie, Marta Rukoz, “TQoS-Transactional and QoS-Aware Selection
Algorithm for Automatic Web Service Composition”, IEEE Transactions on Services Computing,
vol. 3, issue1, pp.73-85, 2010.
[20] Mohamed Sellami, Samir Tata, ZakariaMaamar, and Bruno Defude,”Recommender System for Web
Services Discovery in a Distributed Registry Environment”. Proc. Fourth International Conference on
Internet and Web Applications and Services, pp.418 - 423, IEEE, 2009.
[21] Marwa F. Mohamed, Hany F, ElYamany, Mohamed K. Hussien, Nashwa, M. Yhiea2 and Hamed M.
Nassar,”An Adaptive Replication Framework For Improving The QoS Of Web Services”, vol. 2,
issue 1,SpringerPlus ,2013.
[22] Mario Bravetti, Stephen Gilmore, Claudio Guidi , and Mirco Tribastone, ” Replicating Web Services
for Scalability”, Proc. of Third Symposium on Trustworthy Global Computing Lecture Notes in
Computer Science, Springer, Volume 4912, 2008, pp. 204-221.
[23] Ivona Brandic, Sabri Pllana, Siegfried Benkner, “High-level composition of QoS-aware Grid
workflows: An approach that considers location affinity”, Proc. Workflows In Support Of Large-
Scale Science, pp. 1 - 10, IEEE, 2006.
[24] R.S. Pandey, B.D. Chaudhary, “An estimation of min-max of QoS attributes of a choreography”, In
proc. of International Conference on Advances in Engineering, Science and Management , pp. 872 -
878 ,IEEE ,2012 .

26
[25] Jun Sun, Jin Song Dong,”Design Synthesis from Interaction and State-Based Specifications “,
Transactions on Software Engineering, vol. 32, issue 6, pp. 349 - 364 IEEE, 2006.
[26] An Liu, Qing Li, Senior Member, IEEE, Liusheng Huang, and Mingjun Xiao, “FACTS: A
Framework for Fault-Tolerant Composition of Transactional Web Services”, IEEE Transactions On
Services Computing, vol. 3, issue 1, pp. 46 - 59, 2010.
[27] Johannes Behl, Rüdiger Kapitza, Matthias Schunter , “Providing Fault-tolerant Execution of Web-
service–based Workflows within Clouds “, Proc. 2nd International Workshop on Cloud Computing
Platforms. Bern, Switzerland, article no. 7, ACM, 2012.
[28] Pat. P. W. Chan, Michael R. Lyu,” Reliable Web Services: Methodology, Experiment and Modeling”,
Proc. IEEE International Conference on Web Services, pp. 679 - 686 2007.
[29] Huiyuan Zheng, WeiliangZhao, Jian Yang , Bouguettaya, A.,” QoS Analysis for Web Service
Compositions with Complex Structures” , IEEE Transactions on Services Computing, vol. 6,issue 3,
pp.373 – 386,2007.
AUTHORS
Dr. R. S. Pandey PhD. (Computer Science and Engineering) from Motilal National Institute of
Technology (MNNIT). M.Tech. (Computer Science and Engineering) from Motilal National Institute of
Technology (MNNIT). M.Sc (Computer Science) from J.K Institute of Applied Physics (Allaahabad
University) . Working as Assitant Professor at Birla Institute of Technology since 2000.
Richa Pathak M.Tech.(Computer Science and Engineering ) from Birla Institute of Technology.MCA
from U.P. technical University.Worked as Lecturer at Shri Ramswaroop Memorial College of Engineering
and Technology,Lucknow (U.P.Technical University) from 2010 to 2013.

AN ADAPTIVE APPROACH FOR DYNAMIC RECOVERY DECISIONS IN WEB SERVICE COMPOSITION USING SPACE BASED QOS FACTOR

More Related Content

What's hot (15)

Similar to AN ADAPTIVE APPROACH FOR DYNAMIC RECOVERY DECISIONS IN WEB SERVICE COMPOSITION USING SPACE BASED QOS FACTOR (20)

Recently uploaded (20)

AN ADAPTIVE APPROACH FOR DYNAMIC RECOVERY DECISIONS IN WEB SERVICE COMPOSITION USING SPACE BASED QOS FACTOR