A Broker-based Framework for Integrated SLA-Aware SaaS Provisioning

International Journal on Cloud Computing: Services and Architecture (IJCCSA) Vol. 6, No. 2, April 2016
DOI: 10.5121/ijccsa.2016.6201 1
A BROKER-BASED FRAMEWORK FOR INTEGRATED
SLA-AWARE SAAS PROVISIONING
Elarbi Badidi
College of Information Technology, United Arab Emirates University, Al-Ain, United
Arab Emirates
ABSTRACT
In the service landscape, the issues of service selection, negotiation of Service Level Agreements (SLA), and
SLA-compliance monitoring have typically been used in separate and disparate ways, which affect the
quality of the services that consumers obtain from their providers. In this work, we propose a broker-based
framework to deal with these concerns in an integrated mannerfor Software as a Service (SaaS)
provisioning. The SaaS Broker selects a suitable SaaS provider on behalf of the service consumer by using
a utility-driven selection algorithm that ranks the QoS offerings of potential SaaS providers. Then, it
negotiates the SLA terms with that provider based on the quality requirements of the service consumer. The
monitoring infrastructure observes SLA-compliance during service delivery by using measurements
obtained from third-party monitoring services. We also define a utility-based bargaining decision model
that allows the service consumer to express her sensitivity for each of the negotiated quality attributes and
to evaluate the SaaS provider offer in each round of negotiation. A use-case with few quality attributes and
their respective utility functions illustrates the approach.
KEYWORDS
Cloud computing; Quality-of-Service; Key Performance Indicators; Utility function; SLA monitoring;
Brokerage service
1. INTRODUCTION
Over the last few years, Software as a Service (SaaS) has emerged as one of the most promising
service delivery models in cloud computing. The number of SaaS services in various business
domains grows continually, and new SaaS service offerings emerge on a steady basis. SaaS is
becoming an accepted delivery model for many enterprise applications, including accounting,
collaboration, customer relationship management, enterprise resource planning, human resources
management, etc. As SaaS services proliferate, users and organizations are becoming more
demanding when consuming such services. They increasingly demand services that meet their
functional and non-functional requirements. Therefore, SaaS providers need to negotiate Service
Level Agreements (SLAs) with their service consumers and adhere to their service level
commitments if they want to remain competitive in a changing and demanding business
environment.
Given the variety of consumers’ requirements, SaaS providers need to manage an increasing
number of SLAs, all with potentially changing quality requirements. The SLA lifecycle, as
described in the SLA Management Handbook [24], includes six main phases: SLA Establishment,
Negotiation and Sales, Implementation, Execution, Assessment, and Decommissioning.
Initiatives in standardizing the use of SLAs are (1) WSLA [11] by IBM, (2) the WS-Agreement
specification [1] by the Open Grid Forum, and (3) the SLAng Specification Language [14].
Several efforts-- that we discuss in the Related Work section-- investigated the issue of SLA
Negotiation [8] [10] [11] [22] [25]. This work shares with these efforts the common goal of

2
providing support for automated SLA negotiation and management. Finding the right SaaS
provider is a daunting task for consumers given the abundance and the variety of SaaS offerings.
In the service landscape, the critical issues of service-based systems include service discovery,
service selection, SLA negotiation, and SLA-compliance monitoring. These concerns were
investigated separately using different approaches. We propose in this work an integrated
framework to deal with these issues in the context of a SaaS environment. Its principal
components are the SaaS Broker and the Monitoring Infrastructure. The SaaS Broker 1) mediates
between service consumers and SaaS providers, 2) selects suitable SaaS providers capable of
delivering required functionality and quality-of-service, 3) negotiates SLAs on behalf of service
consumers, and 4) assesses compliance of service delivery with agreed upon SLA. In each round
of negotiation, the SaaS Broker and the selected SaaS provider bargain on multiple SLA
parameters by trying to maximize their global utility functions. The Monitoring Infrastructure is
in charge of observing service delivery, during SLA implementation, using measurements
obtained from independent third party monitoring services.
The remainder of the paper is organized as follows. Next section presents background
information on the issues of negotiation and Service Level Agreements in cloud computing.
Section 3 highlights related work on SLA negotiation in the context of the Service Oriented
Architecture (SOA) and cloud computing environments. Section 4 presents the proposed
framework, describes its components and describes our proposed algorithm for QoS-aware SaaS
selection and our proposed SLA negotiation model. Section 5 presents a SaaS project
management scenario that illustrates the SaaS selection and SLA negotiation. Section 6 concludes
the paper and highlights future work.
2. BACKGROUND
2.1. NEGOTIATION
Negotiation is an important process commonly used in business and personal life to solve
conflicts. It aims at reaching a mutual agreement. For instance, negotiations between unions and
employers aim at reaching a deal regarding salaries and benefits. It is defined in Wikipedia [30]
as: “Negotiation is a dialogue between two or more people or parties intended to reach a
mutually beneficial outcome”.
Negotiation can take different forms and adopt different strategies depending on the multiplicity
of the participating parties:
• One-to-one: This is the most prevalenttype of negotiation, in which a consumer bargains with
a provider for the acquisition of an item or the delivery of a service.
• One-to-many: This is a less familiar form of negotiation, in which a consumer bargains with
several businesses or providers. These providers can compete or cooperate to share the
service delivery to the consumer. An example is when the government negotiates with several
telecom operators for mobile service provisioning.
• Many-to-one: This is also a less common form of negotiation. In this case, several consumers
bargain with a single provider or business. They may compete to reach a deal with the
provider or cooperate to share the item or service offered by the provider.
• Many-to-many: This form of negotiation involves several consumers and several providers or
businesses. A typical example is a negotiation between unions and employers.
Companiescooperate to reach an integrated agreement that saves their interests and unions
bargain to get the best benefits from employers.

3
The simplest form of negotiation is a one-to-one negotiation where the two parties bargain on a
single issue [15]. However, in practice, the two sides often need to negotiate several issues. For
example, a car buyer needs to negotiate with the auto dealer the price, the color, the warranty, the
safety options, and the infotainment options. Multi-issue negotiation is notably more complex
than single-issue negotiation as a human decision relies on her preference on all issues [13]. A
win-win negotiation, in this case, aims to reach a deal that is acceptable to both parties and lets
them feel that they have won. An established way to make the negotiation manageable is first to
characterize the preferences of each participant with a utility function, and then both parties make
decisions based on the evaluation of their respective utility functions. With the proliferation of e-
commerce and electronic transactions, automated negotiation is being increasingly used in
different multi-agent domains including network bandwidth allocation, robotics, space
applications, etc. In these domains, a software agent acts on behalf of a user and negotiates an
SLA with a service provider [4] [3].The negotiation process involves three main components:
• Negotiation objects: The set of issues the negotiating parties negotiate to reach a mutual
agreement. These can include, for example in the case of service provisioning negotiation, the
QoS attributes such as availability, response time, reliability, price, and so on.
• Negotiation protocol: It determines the cardinality of the participating parties, states their
roles, and includes the rules that govern the interaction between the parties. It specifies the
negotiation states, the events that cause the change in those states, and the legal actions of the
participants in each state. Negotiation protocols fall in general into two classes: bilateral
negotiations and auctions. Bilateral negotiations involve two parties, a service provider and a
consumer, and protocol for submitting offers and counter-offers until the two sides reach a
mutually acceptable agreement on the terms and conditions of business, or one of them
withdraws from the negotiation.
• Decision model: the decision-making tool that the negotiating parties use to compute their
negotiation moves and act in line with the negotiation protocol to achieve their objectives.
2.2. SAAS SERVICE LEVEL AGREEMENTS
A Service Level Agreement is anagreement between a service provider and a service consumer
consisting of sections concerning the guarantees and commitments to service levels that the
service provider will deliver. It describes common understandings and expectations of service
between the two parties. The typical components of an SLA are parties, activation-time, scope,
service-level objectives (SLOs), penalties, exclusions, and methods to assess the SLOs [8]. SLOs
represent the goals of the service provider, such as the percentage of service requests the service
provider want handled in a specified number of seconds (e.g., 95% within 10 seconds). They
represent a commitment of the service provider to maintain a particular level of the service in a
predefined period. A typical SLA may have the following SLOs: service availability, system
response time, service outage resolution time, and the reason for an outage.
Every cloud provider uses its key performance indicators (KPIs). There is no standard at this level
yet even though there are some efforts with this regards by the Cloud Services Measurement
Initiative Consortium (CSMIC) [21]. Meegan et al. [18] consider that service consumers should
expect to have general SaaS indicators in their SLA: cumulative monthly downtime, response
time, the persistence of consumer information, and automatic scalability. Burkon L. [2]
summarizes the SLA dimensions and their associated quality indicators for the SaaS delivery
model. These dimensions areAvailability, Response time, Throughput, Timeliness, Reliability,
Scalability, Security, Uptime History, Granularity, Elasticity, Interoperability, Usability,
andTestability.

4
3. RELATED WORK
SLA negotiation and specification of machine-readable SLAs have been the subjects of several
efforts in the context of SOA-based environments, computational grid settings, and recently
cloud-based environments.
Dan et al. [8] described a framework for providing differentiated service levels to service
consumers in an SOA environment using SLAs and automated management. The framework
encompasses WSLA – for the creation and negotiation of SLAs --, a system for dynamic
allocation of resources based on SLOs, a workload management system that orders requests
consistent with the corresponding SLAs, and a system to monitor compliance with the SLA.
Chieng et al. [5] described an SLA-driven service provisioning architecture that allows flexible
SLA negotiation of services. The emphasis is on bandwidth reservation, which is the most
important factor to affect connection’s QoS. The negotiation high-level service parameters are
price, starting time, session length, and guaranteed bandwidth. Di Modica et al. [9] focused on the
use of WS-Agreement for the specification of SLAs and proposed to improve their approach by
adding new functionality to the protocol to allow the parties to an agreement to renegotiate and
adjust its terms during the service provision.
The ambitious goal of several efforts is to support the automated negotiation of the conditions
specified in the SLA. Silaghi et al. [22] introduced a framework for building SLA automatic
negotiation strategies under time constraints in computational grids. The framework relies on the
recent results from negotiation based on learning opponent strategies that agent-based systems
use. They extended the Bayesian learning agent to deal with the limited duration of a bargaining
session and showed that opponent learning strategies lead to satisfaction of participants and an
optimal allocation of resources. Resinas et al. [20] tackled the problem of building automated
service agreement negotiation systems that rely on a bargaining protocol and work in open
environments. They proposed a bargaining architecture that provides support for many
requirements, which are missing from other proposals, such as multi-term negotiation,
heterogeneity of the parties, management of partial information about the parties, and
simultaneous negotiations with different parties. Hasselmeyer et al. [10] described a brokered-
based approach to SLA negotiation. Their solution relies on the idea of outsourcing SLA
negotiation to third parties (agent brokers) acting on behalf of their clients.
An essential component of the European project mOSAIC is the Cloud Agency. The project aims
at enabling service consumers to delegate all management tasks, such as SLA management and
monitoring of the resources to the agency [28]. The Cloud Agency framework includes several
agents that are in charge of managing cloud resources and services provided by various cloud
providers.
The SLA@SOI project aims at defining a comprehensive view of SLAs’ management and
developing a framework for SLA management. A Service Oriented Infrastructure (SOI) can
integrate the framework to manage SLA activities [23]. Theilmann et al. [25] developed, in the
context of this project, a reference architecture for multi-Level SLA management. This
architecture objective is to provide a generic solution for SLA management that can i) cover the
complete SLA and service life cycle; ii) support multiple layers SLA management in a service-
oriented infrastructure, and iii) be used in various use cases and industrial domains.
The CONTRAIL Project proposed and implemented a cloud federation framework, which aims at
relieving the user from managing access to individual service providers [7]. Its federation model,
which relies on SLAs, aims at coordinating management and deployment of applications on
multiple clouds. Each application could then exploit multiple cloud providers. The user negotiates

5
an SLA with the federation, and the federation negotiates SLAs with one or several providers on
behalf of the user. The federation acts as a broker of providers. Selection of providers is SLA-
based. Furthermore, SLA specification in Contrail relies on the SLA@SOI syntax.
Macias et al. [16][17] consider a cloud services market in which a client starts negotiating service
provisioning with the providers that fulfill its requirements. The client creates an SLA proposal
for each provider by filling an SLA template with its requirements. Upon receiving the SLA
proposal from the client, each provider evaluates the proposal and returns its SLA offer with
pricing information and terms of violations. Afterwards, the client selects the provider whose
SLA offer suits best its requirements.
Wu et al. [32] proposed a SaaS broker-based framework for automated SLA negotiation with
multiple SaaS providers. SLA bargaining aims to meet customer needs and relies on some
heuristics and strategies for generating counter proposals to the SaaS provider’s offers. These
heuristics take into account some constraints such as market constraints, time and the trade-off
between QoS parameters. This work and ours share the same goals and objectives.
Our work shares with these efforts the common goal of providing support for automated SLA
negotiation and management. The SaaS Broker with its know-how and value-added services can
assist service consumers to (a) express their service requirements (functional and nonfunctional)
(b) select appropriate SaaS offerings, (c) negotiate SLA terms, and (d) monitor the
implementation of SLAs. However, our approach differs from many of the above approaches,
which select a suitable provider after having conducted SLA negotiation with several providers.
We consider that the selection of potential SaaS providers should be led first, based on the most
recent knowledge on the providers’ offerings, and then SLA negotiation has to be carried out only
with selected SaaS providers in several rounds of proposals and counter-proposals. We also
believe that SLA compliance monitoring should be carried out in collaboration with third-party
monitoring services to guarantee the independence of metrics’ measurement.
4. METHODS
4.1.FRAMEWORK OVERVIEW
Figure 1 depicts our proposed integrated framework for SaaS provisioning. The components of
the framework are service consumers, the SaaS Broker, the monitoring infrastructure, and SaaS
providers.
4.1.1.SAAS BROKER
The SaaS Broker is a mediator service that decouples service consumers from SaaS providers. It
allows users to subscribe to some services and enables SaaS providers to market their service
offerings. Given that service providers and consumers want to focus more on their core
businesses rather than negotiating, managing, and monitoring QoS, they delegate management
tasks, such as service selection and SLA negotiation, to the SaaS Broker.
The SaaS Broker offers the following management operations: Identity and Access Management,
QoS-based SaaSSelection, SLA Negotiation, and Policies Management. Several components
implement these operations and cooperate to deliver personalized services to service consumers
and SaaS providers. Figure 2 depicts the architecture of the SaaS Broker, which includes the
following components: theCoordinator, the Profile Manager, the Selection Manager, the SLA
Manager, the Policy Manager, and the SLA Monitoring Manager. The back-end databases

6
maintain information about profiles and preferences of service consumers, SaaS providers’
policies, SLAs, and dynamic QoS information. The Profile Manager manages the profiles of
service consumers, including their preferences for personalized services and QoS. The Selection
Manager is in charge of implementing policies and strategies for the selection of suitable SaaS
providers, based on the service consumer’s functional and non-functional requirements and the
SaaS providers’ QoS offerings. The SLA Manager is in charge of carrying out the negotiation
process between a service consumer and a SaaS provider to reach agreement as to the service
terms and conditions. The Policy Manager manages different kinds of policies such as
authorization policies, and policies for monitoring services and their QoS. Specific requirements
or capabilities of a service provider are declared using XML policy assertion elements. Each
assertion describes an atomic aspect of service requirements (e.g. authentication scheme,
authorization scheme, QoS characteristics).
Figure 1. Framework for SLA-driven service provisioning
4.1.2.THE MONITORING INFRASTRUCTURE
SaaS services work well most of the time. But, it does not mean that outages and performance
issues never happen. Most SaaS providers guarantee availability levels of 99.9% uptime, which
translates into about 43 minutes of downtime per month that one can consider as not too much
downtime.However, the more a service consumer uses the SaaS service extensively, the more any
disruption in service delivery is amplified. The consequences of an interruption in the middle of
an important business transaction could be costly.
Many SaaS providers use a “Trust” website, which provides information on service status and
disruptions to facilitate accountability for service provision. However, the cause of any disruption
in service delivery often comes later after an in-depth assessment of the situation has been made
and posted publicly.
Repeated outages or slowdowns might lead service consumers to complain or go away trying to
find other alternative providers. With an agreement on a proactive approach to monitoring in real
time of the SaaS delivery, the provider can quickly make corrections to mitigate disruptions.
Independent performance monitoring from a third-party helps the provider put the required

7
resources to keep the service running at its full potential. It also helps the service consumer
understand the level of service and performance obtained from the provider, and, therefore
optimize its business processes the SaaS application is supporting. An example of a third-party
monitoring service is Keynote [12] with its solution for “Web monitoring for SaaS
applications”.Other monitoring services includeMonitis[19] and Uptrends [27].
As depicted in Figure 2, the Monitoring Infrastructure includes Monitoring Plug-insthat allow
interacting withthird-party Monitoring Services. Each Monitoring Plug-in is in charge of mapping
resource metrics, measured by the Monitoring Service, into SLA parameters and monitoring
current service levels and their compliance with SLA. Monitoring Plug-ins may be added to the
framework if service consumers and SaaS providers agree on using the services of particular
monitoring companies. Plug-ins typically use APIs provided by Monitoring Services to get
service measurement data. For example, Monitis is providing an open API that allows accessing
most of the commands that are available from Monitis dashboard. Similarly, Keynote is offering
an API to access Keynote measurement data.
Figure 2. Architecture of the SaaS Broker and the monitoring infrastructure
4.2.SLA-BASED SERVICE PROVISIONING
Service provisioning is typically carried out in four phases: Expression of the consumer’s
requirements and service level expectations, selection of a potential SaaS provider, SLA
negotiation and finalization of the contract, service delivery,and SLA compliance monitoring.
4.2.1. EXPRESSION OF THE CONSUMER REQUIREMENTS
In this phase, the service consumer submits an SLA Request to the SaaS Broker seeking an
appropriate SaaS provider that meets its service functional and non-functional requirements. The
request describes the level of service that the service consumer is willing to accept and includes

8
some QoS indicators such as response time, availability, and throughput. After processing the
service consumer’s authentication, the Coordinator requests its profile from the Profile Manager.
If the service consumer’s profile is not available in the profile repository, then the Coordinator
asks the service consumer to provide additional information, such as service preferences and
desired levels of service, to create a new profile for the service consumer. Figure 3 illustrates an
example of quality requirements for a project management SaaS service.
Figure 3. An example of quality requirements specification
4.2.2. QoS-aware Service Selection
Before starting the SLA negotiation, the Coordinator requests from the Selection Manager to
select appropriate SaaS providers based on the user’s QoS requirements and the QoS offerings of
SaaS providers advertised with the SaaS Broker or a public registry.
To enter the market, SaaS providers often advertise their service offerings using templates that
express the functional and non-functional characteristics of their services. Not all SaaS providers
are created equal. They vary by the vertical they are serving, by their maturity, and by the type of
application they are offering to their customers. Therefore, service offerings can vary widely
regarding service levels and QoS they can deliver. Service descriptions may also vary as they
may use different description languages such as WSDL (Web services Description Language) [6],
WSOL (Web Services Offering Language) [26], or proprietary languages. The description of their
QoS offerings may also be heterogeneous. Various approaches have been used to add QoS
support to Web services’ description. The first approach advocates incorporating QoS parameters,
such as response time, throughput and cost, into the WSDL service description. Other works
propose extensions to the Web services Policy Framework to represent QoS policies of Web
services. With this heterogeneity in service descriptions, it would be difficult to match services
offerings to the user’s needs without using a common standard description model (such as WSDL
model with QoS support) to map all service offerings to that common model.
A SaaS provider may exceed the service consumer's expectations for some quality attributes
while its offering might be below the user’s expectations for other quality metrics. Thus, it is
imperative to evaluate the overall offering of every potential SaaS provider by taking into account

9
all quality attributes. Such evaluation will allow ranking SaaS providers and selecting the most
promising one that can satisfy the consumer request. We describe in the following our proposed
algorithm that takes into consideration the QoS requirements of the service consumer and the
knowledge of the SaaS Broker on the various service offerings.
Let = { , , … , } be the list of QoS attributes, such as availability, throughput, response
time, reputation, and cost of service, required by the service consumer. We classify these QoS
attributes into two categories. The first category includes utility-driven QoS attributessuch as
availability and throughput that service consumers would like to maximize. The second category
includes cost-driven QoS attributes such as response time and cost of the service that service
consumers would like to minimize.
Let = { , , … , } be the list of potential SaaS providers, which are capable of
providing the type of service requested by the consumer. Let = { , , … , }, be the vector of
quality requirements of the service consumer for each QoS attribute in Q.The following vector
expresses the QoS offering of the SaaS providerSP .
= { , , … , }
The above QoS attributes are expressed in different units and cover a range of values. To be able
to sort and rank the SaaS providers’ offerings regarding their QoS offerings, we need to
normalize values of the QoS offers and define utility functions that map the vector of QoS values
into a single real value. We define:
= max
= min
as the maximum and minimum values of the QoS attribute Qi considering all the offers of the
potential SaaS providers in SP. The requirement of the service consumer for may fall in the
range [ , ] or outside of this range. To take account of the requirements of the service
consumer during the normalization, we define the two following values:
′ = "#$ ( , )
′ = "'(( ), )
We define * , the normalized value for the offer of the SaaS provider regarding the QoS
attribute Q in equation (1) as:
* = ,
-./
012
3-/
4
-./
012
3-./
0/5 , for 9:;< − >?'@A( QoS attributes
GH
I
3G.H
JHK
G.H
JLM
3G.H
JHK , for N<'O'<P − >?'@A( QoS attributes
Q (1)
Let R = {9 , 9 , … , 9 }, with 0 ≤ 9 ≤ 1, be the vector of normalized values of the quality
requirements that the service consumer. When * ≥ 9 , it means that the offer of SaaS
provider exceeds the expectation of the service consumer regarding the quality attribute .
Otherwise, it means that the offer does not meet the consumer’s expectation for .The SaaS
Broker may assign (on behalf of the service consumer) an importance weight to each quality
attribute.Let W = {X , X , … , X } be the weight vector associated with the quality attributes

10
respectively.We define the Utility function (weighted combined level of satisfaction) associated
with the offer of SaaS provider as:

Y = ∑ X *[ (2)
Equation 2 allows ranking the QoS offers of the potential SaaS providers capable of offering the
functionality required by the service consumer. The best service offer corresponds to the offer
that maximizes the utility function Y .
Best service offer "#$ [ Y (3)
Figure 4 summarizes the different steps of the algorithm.Once the Selection Manager has
selectedthe most promisingSaaS provider, the SaaS Brokerstarts negotiating SLOs with that SaaS
providerthrough its SLA Manager.
Figure 1. QoS-aware selection algorithm
4.2.3. NEGOTIATION MODEL
The SLA negotiation process of the proposed framework is a bargaining protocol where the SLA
Managers of the SaaS Broker and SaaS provider negotiate multiple attributes (QoS attributes).
The steps of the SLA negotiation protocol are as follows:
• Step1: The SLA Manager of the SaaS Broker forwards the SLA Request to the SLA Manager
of the selected SaaS provider requesting an SLA proposal. The SLA request describes only
the user’s required level of service regarding quality indicators. The SaaS provider parses the
SLA Request and validates it against its SLA templates.

11
• Step 2: If the SLA Request is acceptable to the SaaS provider, then its SLA Manager
responds to the SLA Request by sending back an SLA proposal to the SaaS Broker. The SaaS
Brokeranalyses it to find out if it meets or not to all the service consumer’s functional and
non-functional requirements. A decision model is used at this point to evaluate the proposal
of the SaaS provider.
• Step 3: If the SaaS provider can meet the service consumer’s expectations, then the SaaS
Broker accepts the offer of the SaaS provider and sends an SLA Confirmation to the SLA
Manager of the SaaS provider. Otherwise, it rejects the offer and makes a counter-proposal
with different conditions, terms, and cost.
• Step 4: In the case of the SLA Confirmation, the two parties, service consumer and SaaS
provider, approve the agreement, and the service consumer can start using the service
according to the terms of the agreement. The agreement specifies the service that the provider
should offer to the service consumer, the activation time, the level of QoS to guarantee, the
cost of service, the validity period, and the actions to take in the case of a violation of the
agreement.
In this process, a multi-issues negotiation takes place between the SLA Managers of both parties,
as they have to negotiate multiple QoS attributes concurrently. To reach agreement, The SLA
Managers of both parties have to go through several rounds of negotiation of offers and counter-
offers until they reach agreement or reach a predefined maximum number of rounds. Figure 5
shows the state charts of the SLA Managers of the negotiating parties. The SaaS Broker, upon
reception of an SLA request, maps the service consumer's expectations into normalized QoS
values in the range]0,1^.
In each round of negotiation, as illustrated in Figure 5, the SLA Manager of the SaaS Broker
evaluates the utility function of each QoS attribute and the global utility function to determine
whether the offer of the SaaS provider is acceptable or not.Various functions may be used to
express the service consumer’s utility for eachquality attribute . If is a utility-driven attribute,
such as the availability, its utility function is at its maximal when the SaaS provider can ensure
100% for that quality. If is a cost-driven attribute, such asresponse time, the corresponding
utility function is at its maximal when the SaaS provider can guarantee a lower value, which is
very close to zero, for that quality. Let V be the set of utility-driven quality attribute from Q; and
Let R be the set of cost-driven quality attribute from Q. = _ ∪ a.
We adopt in Equation (4)the following function b to express the utility function of a utility-driven
quality attribute X:
b($) =
cd ( efd)
efd
cd
(4)
$ is the normalized value of the offer made by the SaaS providerfor X. gh is the SLA value for X.
ihrepresents the sensitivity of the service consumerwith respect to the quality attribute X. When
ih is equal to zero, the service consumer is indifferent to the quality attribute X. When ihis
equal to one, the service consumer is moderately sensitive to the quality attribute X (the
relationship is linear). When ihis greater than one, the service consumer is increasingly sensitive
to the quality attribute X. Whenih increases, the service consumer is expressing increasing
concern about it. For values of ih smaller than one, the service consumer is expressing increasing
indifference for the quality attribute X when ihdecreases to reachzero.
Similarly, we consider in Equation (5)the function jthat expresses the utility function of a cost-
driven quality attribute Y:

12
j(P) =
3kcl
eflkcl
(5)
y is the normalized value of the offer made by the SaaS provider for the quality attribute Y. gm is
the SLA value for Y. imrepresents the sensitivity of the service consumer toward the quality
attribute Y.jreaches its maximum, which is 1, when y is 0 and decreases to 0 when y reaches 1.
(a) SLA Manager of the SaaS Broker
(b) SLA Manager of the SaaS Provider
Figure 5. State chart of the SLA Managers of the negotiating parties.
We define in Equation (6) U as the global utility function, which serves to evaluate the SaaS
provider offer during the negotiation phase. We assume that the QoS attributes are independent.
Therefore, U can be expressed by the additive linear function of the individual utility functions (F
and G) of the quality attributes . 1 ≤ ' ≤ (
Y =
∑ X b
'
∈ _
+
∑ X j
p
∈ a
(6)
b represents the utility function associated with the utility-driven QoS attribute . j represents
the utility function associated with the cost-driven QoS attribute .X is the weight that the

13
service consumer assigns to that attribute . Each weight is a number in the range [0,1]
and∑ X = 1.
5. RESULTS
In this section, we consider a scenario from an application domain in which many SaaS offerings
are available to businesses. The scenario is used to illustrate the selection process and the
negotiation model.
A business company that we call COMPANY would like to enhance its productivity by using
project management solutions. On-premise project management tools have limitations when it
comes to the mobility of the workforce and the variety of devices that employees may use ranging
from desktops to tablets and smartphones. Therefore, a cloud-based solution for project
management is increasingly becoming attracting given the way teams may communicate and
collaborate, the benefits, and the flexibility it offers to businesses.
Indeed, today businesses have many SaaS solutions for project management support. Although
the solutions have many similarities, each solution has its unique features, strengths, and
weaknesses. SaaS services for project management are changing the way businesses manage
projects. Managers no longer have to print or email on a weekly basis Gantt charts to team
members to update them on project status. Besides, the management of multiple projects does not
require anymore tracking and exchanging multiple Microsoft Project documents. Examples of
SaaS solutions for project management are basecamp, GroupCamp, Planzone, Clarizen, Zoho
project, and Beesy. Table 1 presents the main features of these SaaS offerings.
Table 1- SaaS Project Management Solutions
Operation SaaS PM
solution
basecamp GroupCamp Planzone Clarizen Zoho
project
Beesy
Defining tasks ✓ ✓ ✓ ✓ ✓ ✓
Assigning resources ✓ ✓ ✓ ✓ ✓ ✓
Specifying milestones ✓ ✓ ✓ ✓ ✓ ✓
Generating detailed
reports
✓ ✓ ✓ ✓ ✓ ✓
Parallel tasks support ✓
Sending email to
create new task
✓ ✓ ✓ ✓ ✓
Templates to specify
project data
structures (team
members, tasks, wiki
page, etc..)
✓ ✓ ✓
Advanced reporting ✓ ✓ ✓
Pricing scheme Monthly
subscription
& yearly
subscription
Six pricing
plans: (Free,
Start, Plus,
Best, Max)
Four
monthly
plans
(Basic,
Team,
Business,
Enterprise)
&
customized
plan.
Three
Monthly
plans (
Professional/
Enterprise/
Unlimited)
Four
monthly
plans (Free,
Express,
Premium,
Enterprise)
Four
monthly
plans
(Free,
Sync, Pro,
Team)
SLA support Yes for
yearly
subscription
SLA
management

14
Selecting a SaaS solution by COMPANY is not an easy task given the plethora of features,
pricing schemes, and SLAs. COMPANY has first to understand and specify its project
management needs. Then, it can delegate the selection of an appropriate SaaS solution to the SaaS
Broker, who has better knowledge on the features of each SaaS solution. The selection process
starts by considering the solutions that might satisfy the functional requirements of COMPANY.
Then, potential SaaS offerings will be ranked based on its non-functional requirements. In this
first phase, high-level requirements, such as the scope of COMPANY’s project management
needs, might be used to identify candidate services. The scope may be project specific,
department level, or enterprise level. The selection process might then be refined with other
detailed or low-level functional requirements. Assuming that the SaaS Broker has knowledge on
the non-functional offerings of the potential SaaS providers, our proposed algorithm might be
used to rank them.
We assume that the QoS requirement of the service consumer are: Availability=99.97%,
Reliability=99.96%, Cost=25$, and Response time=6ms. Table 2 shows the QoS values offered
by 24 potential SaaS providers. Table 3 lists the weights that the SaaS Broker assigns to these
QoS parameters according to the consumer requirements.By using our selection algorithm,
normalized values are calculated for each QoS parameter while differentiating between utility-
driven QoS attributes and cost-driven QoS attributes. Then, using the weight table, the aggregate
utility function is calculated for each SaaS offering. Figure 6 depicts the normalized values of
each of these QoS attributes.
Table 2- Qos Offerings of SaaS Providers
SaaS_ID Availability Reliability Cost ($) Response time (ms)
1 0.99988 0.9995 16.1 6
2 0.99968 0.99953 38.1 2
3 0.99935 0.99962 8.4 3
4 0.99988 0.99964 40.2 3
5 0.99959 0.99954 12.6 4
6 0.99963 0.99958 22.2 6
7 0.99939 0.99971 33.2 7
8 0.99918 0.99975 25.3 2
9 0.99995 0.9999 30.8 7
10 0.99958 0.99956 24.2 3
11 0.99945 0.99971 22.8 7
12 0.99981 0.99976 6.7 7
13 0.99911 0.99987 15 3
14 0.99924 0.99983 11.7 5
15 0.99912 0.9998 24.8 5
16 0.99948 0.99973 22.7 3
17 0.99952 0.99967 31.9 5
18 0.99999 0.99962 23.9 2
19 0.99944 0.99975 33.7 3
20 0.99943 0.99972 20.7 5
21 0.99987 0.99957 19.6 5
22 0.99959 0.99977 27.2 2
23 0.9997 0.99952 20.4 4
24 0.99999 0.99992 25.4 5
Table 4 depicts the rank table generated by the SaaS Broker using our proposed algorithm. The
offer of the SaaS provider with ID 24 is the best offer as it maximizes the aggregated utility. In
parallel with that, we also use the TOPSIS method [31] for ranking the SaaS offerings. We
identified first the ideal solution (IS) and the negative ideal solution (NIS). Then, we calculated

15
for each SaaS offer the separation from IS from the NIS. Afterwards, we calculate the closeness
to the ideal solution (CIS). Table 5 depicts the ranking using the TOPSIS method.
Figure 6. Normalized values of the QoS attributes
Table 3- Weighted table of the QoS attributes
Availability (%) Reliability (%) Cost ($) Response time (ms)
Weight 0.305 0.267 0.197 0.231
The results show that both algorithms have obtained the same best SaaS provider (SaaS ID: 24).
Subsequent SaaS offerings in the two rankings are very close, especially for the first ten best
offerings. This process results in identifying a suitable SaaS provider for project management.
Reaching an agreement with that provider would be the next phase before COMPANY would be
able to use the SaaS solution for its project management activities.
To illustrate our negotiation model, we consider that negotiation would be mainly about two
quality attributes: availability and response time. In this scenario, both availability and response
time are important qualities for COMPANY. The desired level of the availability is 99.97%
uptime and 20% (6ms) for the response time (qr= 0.99 and qs= 0.20). In addition, COMPANY
gives 70% importance weight to the availability and 30% weight to response time (tr= 0.70 and
ts= 0.30).
Figure 7 shows the evolution of the utility function associated with Availability for different
values of βv, the sensitivity of COMPANY to the availability of the service (iw= 1, 2, and 4
respectively). At low offers of availability by the SaaS provider, the utility is also low. As the
SaaS offer increases, the utility rapidly increases when COMPANY is not sensitive to the
availability (iw= 1), moderately when iw= 2, and slowly as availability becomes a sensitive issue
for the business (iw= 4).

16
Table 4- Ranking of SaaS Offerings using our
proposed algorithm
SaaS_ID Utility (%) Rank
24 65.90 1
12 60.49 2
9 60.07 3
18 47.71 4
21 42.90 5
14 42.24 6
22 41.45 7
1 40.86 8
13 38.34 9
16 37.74 10
20 36.54 11
4 35.59 12
5 35.41 13
11 35.37 14
3 34.65 15
6 33.69 16
23 33.36 17
19 31.15 18
17 29.90 19
10 29.51 20
15 28.47 21
7 27.17 22
8 27.08 23
2 22.90 24
Table 5- Ranking of SaaS Offerings using the
TOPSIS method [31]
SaaS_ID CIS (%) Rank
24 70.60 1
18 65.05 2
22 62.23 3
9 58.97 4
12 57.62 5
4 55.29 6
16 54.66 7
21 52.70 8
13 51.54 9
19 49.76 10
3 49.12 11
14 48.81 12
2 48.22 13
1 47.97 14
8 47.94 15
5 47.91 16
23 47.71 17
10 47.71 18
20 45.03 19
6 40.47 20
17 40.42 21
15 39.25 22
11 37.70 23
7 31.35 24
Figure 7. Utility function of the availability (bw) with different values of the sensitivity factoriw
Similarly, Figure 8 shows the evolution of the utility function associated with response time for
different values of ix, the sensitivity of COMPANY to the response time quality attribute (ix= 1,
2, and 4 respectively). At low response time offers by the SaaS provider, the utility is high. As the
SaaS offer for response time increases, the utility rapidly decreases when COMPANY is not
sensitive to the response time (ix= 1), moderately when ix= 2, and slowly when response time is
a sensitive issue for the business (ix= 4).

International Journal on Cloud Comput
Figure 8. Utility of the response time (
Assuming that COMPANY is more sensitive to availability than to
2).Figure 9 plots the additive utility function for this scenario. It shows that the
function reaches its maximal value when the
is high (close to 1).The SaaS Broker, acting on behalf of COMPANY, may set a threshold for the
overall utility function. When the offer of the SaaS provider is below that threshold, the SaaS
Broker rejects the offer and sends a counter
Figure 9.Overall u
6. CONCLUSION
With the rising adoption of Software
providers need to implement and
highly demanding business environment. On the other hand, service consumers need to find
appropriate SaaS providers that meet their functional and non
negotiate with them the terms of service delivery.
. Utility of the response time (jx) with different values of the sensitivity factor
COMPANY is more sensitive to availability than to response time (yr
plots the additive utility function for this scenario. It shows that the overall
function reaches its maximal value when the response time is low (close to 0), and the
The SaaS Broker, acting on behalf of COMPANY, may set a threshold for the
offer and sends a counter-proposal to the SaaS provider.
Overall utility function of the service consumer
With the rising adoption of Software-as-a-Service as a service provisioning model, SaaS
providers need to implement and manage a growing number of SLAs to remain competitive in a
appropriate SaaS providers that meet their functional and non-functional requirements and
m the terms of service delivery.
ing: Services and Architecture (IJCCSA) Vol. 6, No. 2, April 2016
17
) with different values of the sensitivity factor ix
r= 4 and ys=
overall utility
and the availability
The SaaS Broker, acting on behalf of COMPANY, may set a threshold for the
Service as a service provisioning model, SaaS
manage a growing number of SLAs to remain competitive in a
functional requirements and

18
In this paper, we have presented a framework for SLA-based service provisioning, which relies
on a SaaS Broker for a) mediating between service consumers and SaaS providers, b) selecting
suitable SaaS providers that can meet service consumers’ expectations, and c) negotiating the
terms of the SLA on behalf of service consumers. SLA negotiation involves the negotiation of
multiple QoS attributes with both parties trying to maximize their utility function in several
rounds of proposals and counter-proposals. We proposed a SaaS selection algorithm and a
decision negotiation model that the SaaS Broker may implement to evaluate the offers of the
selected SaaS provider in each round of negotiation. We have considered a scenario with two
quality attributes, availability and response time, to illustrate the approach. The proposed decision
model allows the consumer to specify her sensitivity for each negotiated quality attribute. Our
approach differs from other approaches that carry out SLA negotiation with several cloud service
providers, and then select the one with the best offer. It first selects the best suitable SaaS
provider based on prior knowledge of the SaaS Broker on the providers’ offerings. Then, it
negotiates with the selected SaaS provider the SLA terms in several rounds of proposals and
counter-proposals until an agreement or a timeout is reached.
As future work, we intend to investigate how the SLA negotiation evolves when the importance
weights, associated with the QoS attributes, vary in time and, then, study the effect of the
interdependence between QoS attributes on the SLA negotiation. Also, we intend to build a
prototype of our proposed framework.
REFERENCES
[1] Andrieux, A., et al., (2007) Web services agreement specification (ws-agreement). Technical report,
Open Grid Forum.
[2] Burkon, L., (2013) Quality of Service Attributes for Software as a Service. Journal of Systems
Integration, pp. 38-47.
[3] Chen, Y.M., and Huang, P.-N., (2009) Agent-based bilateral multi-issue negotiation scheme for e-
market transactions. Applied Soft Computing, 9(3), 1057–1067.
[4] Chevaleyre, Y., Dunne, P.E., Endriss, U., Lang, J., Lemaitre, M., Maudet, N., Padget, J., Phelps, S.,
Rodriguez-Aguilar, J.A., and Sousa, P., (2006) Issues in multiagent resource allocation. Informatica,
30, 3–31.
[5] Chieng, D., Marshall, A., and Parr, G., (2005) SLA brokering and bandwidth reservation negotiation
schemes for QoS-aware internet. IEEE Transactions on Network and Service Management, 2(1), 39-
49.
[6] Christensen, E., Curbera, F., Meredith, G., and Weerawarana, S., (2001) Web services description
language (WSDL) 1.1. World Wide Web Consortium (W3C) note, online at: http://
www.w3.org/TR/wsdl, March 15.
[7] Coppola M., Dazzi P., Lazouski A., Martinelli F., Mori P., Jensen J., Johnson I., Kershaw P., (2012)
The Contrail approach to cloud federations. Proc. of the International Symposium on Grids and
Clouds (ISGC’12). http://pos.sissa.it/archive/conferences/153/019/ISGC 2012_019.pdf
[8] Dan, A., Davis, D., Kearney, R., & Keller, A. (2004) Web services on demand: WSLA-driven
automated management. IBM systems Journal, 43(1), 136-158.
[9] DiModica, G., Tomarchio, O., & Vita, L., (2009) Dynamic SLAs Management in Service Oriented
Environments. Journal of Systems and Software, 82(5), 759-771.
[10] Hasselmeyer, P., Qu, C., Schubert, L., Koller, B., & Wieder, P., (2006) Towards Autonomous
Brokered SLA Negotiation. Exploiting the Knowledge Economy: Issues, Applications, Case Studies,
IOS Press, Amsterdam.
[11] Keller, A., & Ludwig, H., (2003) The WSLA Framework: Specifying and Monitoring Service Level
Agreements for Web Services, Journal of Network and Systems Management, 11(1), 57-81.
[12] Keynote systems (2014) Mobile and Web Performance Testing, Monitoring and Analytics. From
http://www.keynote.com
[13] Lai, G.M., Li, C.H., and Sycara, K., (2006) Efficient multi-attribute negotiation with incomplete
information. Group Decision and Negotiation, 15 (5), 511–528.

19
[14] Lamanna, D., Skene, J., & Emmerich, W., (2003) SLAng: A Language for Defining Service Level
Agreements. Proc. ofthe Ninth IEEE Workshop on Future Trends of Distributed Computing Systems,
28-30 May, pp. 100–106.
[15] Lee, C.C., Ferguson, M.J., (2010) To reveal or not to reveal? Strategic disclosure of private
information in negotiation, European Journal of Operational Research,207 (1), 380–390.
[16] Macias, M., Fito, O., Guitart, J., (2010) Rule-based SLA management for revenue maximisation in
cloud computing markets, in: 2010 Intl. Conf. of Network and Service Management, CNSM’10,
Niagara Falls, Canada, pp. 354–357.
[17] Macias, M., Guitart, J., (2010) Using resource-level information into nonadditive negotiation models
for cloud market environments, in: 12th IEEE/IFIP Network Operations and Management
Symposium, NOMS’10, Osaka, Japan, pp. 325–332.
[18] Meegan, J., et al., (2012) Practical Guide to Cloud Service Level Agreements, Version 1.0. Cloud
Standards Customer Council.
[19] Monitis (2014) Network & IT Systrems Monitoring - Monitor Everything. From
http://www.monitis.com/
[20] Resinas, M., Fernández, P., & Corchuelo, R., (2012) A bargaining-specific architecture for supporting
automated service agreement negotiation systems. Science of Computer Programming, 77(1), 4-28.
[21] Siegel, J.; Perdue, J. (2012) Cloud Services Measures for Global Use: The Service Measurement
Index (SMI). Proc. of the Annual SRII Global Conference (SRII), 24-27 July, pp. 411-415.
[22] Silaghi, G. C., Şerban, L. D., & Litan, C. M., (2012) A time-constrained SLA negotiation strategy in
competitive computational grids. Future Generation Computer Systems, 28(8), 1303-1315.
[23] SLA@SOI project (2014) Empowering the Service Economy with SLA-aware Infrastructures. from
http://www.sla-at-soi.eu/
[24] TeleManagement Forum, (2004) SLA Management Handbook, TeleManagement Forum Member
Evaluation, Version 2.0.
[25] Theilmann, W., Happe, J., Kotsokalis, C., Edmonds, A., Kearney, K., & Lambea, J., (2010) A
Reference Architecture for Multi-Level SLA Management. Journal of Internet Engineering, 4(1), 289-
298.
[26] Tosic, V., Patel, K., and Pagurek, B., (2003) WSOL—a language for the formal specification of
classes of service for web services. Proc. of the International Conference on Web Services
(ICWS’03), 23-26 June, pp. 375–381.
[27] Uptrends (2014). Website monitoring, server monitoring, monitor uptime by Uptrends. From
http://www.uptrends.com/
[28] Venticinque, S., Aversa, R., Di Martino, B., Rak, M., & Petcu, D., (2011) A cloud agency for sla
negotiation and management. Proc. of the Euro-Par 2010 Parallel Processing Workshops, Vol. 6586
of the series Lecture Notes in Computer Science, 587-594.
[29] Waeldrich, O., et al., (2011) WS-Agreement Negotiation Version 1.0. Open Grid Forum.
[30] Wikipedia, (2016). Negotiation.  https://en.wikipedia.org/wiki/Negotiation
[31] Hwang, C.L., Yoon, K. (1981). Multiple Attribute Decision Making: Methods and Applications. New
York: Springer-Verlag.
[32] Wu, L., Garg,S.K., Buyya, R., Chen, C., and Versteeg, S. (2013). Automated SLA Negotiation
Framework for Cloud Computing. In Proc. of the 13th IEEE/ACM International Symposium on
Cluster, Cloud, and Grid Computing, pp. 235 – 244.

A Broker-based Framework for Integrated SLA-Aware SaaS Provisioning

More Related Content

What's hot (7)

Similar to A Broker-based Framework for Integrated SLA-Aware SaaS Provisioning (20)

More from neirew J (20)

Recently uploaded (20)

A Broker-based Framework for Integrated SLA-Aware SaaS Provisioning