ADDING MEANING TO QOS NEGOTIATION
Cláudia M. F. A. Ribeiro
Universidade do Estado do Rio Grande do Norte, Av. Itapetinga, 1430, Natal, Brazil
Nelson S. Rosa, Paulo R. F.Cunha
Universidade Federal de Pernambuco, Cidade Universitária, Recife, Brazil
Keywords: QoS Negotiation, Semantic Web Services, SLA Ontology.
Abstract: Using quality of service (QoS) to discover Web Services that better meet users’ needs became a key factor
to differentiate similar services. Treating QoS includes negotiating QoS capabilities, since there is a
potential conflict between service provider and service requestor requirements. This paper addresses this
problem by using an ontological approach for QoS negotiation that aims to improve user participation. For
this purpose, a Service Level Agreement (SLA) ontology that explicitly considers subjective user QoS
specification was conceived. Its internal components and the role it plays during service discovery are
detailed.
1 INTRODUCTION
Evidence suggests that, by focusing on quality,
companies can increase market share. In Web
Service market, similar services can be differentiated
by quality of service (QoS) properties, which is a set
of non functional properties that fully characterizes
“how” a service works, instead of “what” it does
(Chung, 2000). In multimedia applications, QoS
providing means to meet user needs covering more
than just price, i.e. other requirements, such as
usability and security, can also be used to identify
the desired service.
Treating QoS includes negotiating QoS
capabilities, since there is potentially conflict
between service provider and service requestor
requirements. Indeed, the service provider aims to
minimize resource allocation, whereas the desired
QoS level defined by user tends to increase resource
usage. While QoS negotiation requires QoS
specification, it is apparent the necessity to include
QoS information in Web Services description.
Today’s Web Services description is mainly
automation oriented, based on application to
application communication. It means that service
description is focused on how to make service
discovery, selection, invocation and monitoring,
more automatic processes. For instance, automatic
discovery involves automatically locating Web
Services that provide a particular service with
appropriate properties (Booth, 2005).
Semantic Web Services are considered the next
generation of Web Services (Alesso, 2004). They
differ from traditional ones by using ontology, such
as OWL-S (Martin, 2003) and WSMO (Feier, 2005),
to add meaning to their description, although they do
not prioritize QoS. Considering that human beings
have used more and more Internet as a preferred
environment, both for document search and for
localizing services, it is essential to consider their
QoS needs, since they can perceive changes in QoS
level actually provided.
In this work, it is presented a model for semantic
QoS negotiation, which is defined as the process of
Web Services selection, considering service QoS
capabilities. It also includes abstract considerations
about resource negotiation for QoS guarantees based
on ontology. For this purpose a domain ontology for
Service Level Agreement (SLA) was conceived.
SLA is also known as QoS contract and may include
QoS specification from different perspectives, from
subjective user level to low level network
mechanisms.
Despite the use of ontologies for QoS, and
specifically for SLA were subject of important
research activities (Zhou, 2005), (Menascé, 2002),
there is a gap in the explicit treatment of human
needs during QoS negotiation. A premise used by
106
M. F. A. Ribeiro C., S. Rosa N. and R. F.Cunha P. (2006).
ADDING MEANING TO QOS NEGOTIATION.
In Proceedings of the Eighth International Conference on Enterprise Information Systems - SAIC, pages 106-111
DOI: 10.5220/0002463401060111
Copyright
c
 SciTePress

this work is that meaningful QoS contracts (Ribeiro,
2004a) can be used to improve the user participation
during QoS negotiation and monitoring, contributing
to user satisfaction. The internal components of SLA
ontology proposal and its role at QoS negotiation are
described.
This paper is structured as follows. Section 2 is
dedicated to related work. Section 3 describes
ontology languages and main ontology concepts,
especially OWL language, which was used during
SLA ontology definition. Section 4 deals with SLA
ontology and its internal components. Section 5
discusses about a scenario where semantic QoS
negotiation is used. Finally, Section 6 presents
conclusions and future work.
2 RELATED WORK
To realize semantic QoS negotiation, as considered
in this work, it is necessary to represent QoS
information at different perspectives. Using
ontology to represent QoS concepts is not a new
approach. The FIPA-QoS ontology (FIPA, 2003)
deals with QoS at FIPA Message Transport Service.
QoS ontology for Workflows and Web Services is
described at (Cardoso, 2004). In general, these
ontologies are conceived for specific purposes and
they usually lack representing of subjective end user
QoS.
SLA ontologies are subject of research activities
as well. IBM’s WSLA (Ludwig, 2003) is a
representative initiative to specify SLA parameters
associated with services and to identify signatory
parties of the contract (provider and customer).
However, as traditional approaches, WSLA was
mainly defined through technical parameters that are
essential for the service configuration and
monitoring, but frequently have no meaning for
typical users limiting their participation during QoS
negotiation. Next section is dedicated to ontology
languages, especially OWL (Dean, 2004), which
was used to describe SLA ontology proposed.
3 WEB ONTOLOGY
LANGUAGES
In computer science, ontology is defined as a
software artefact that allows representing and
sharing of domain specific knowledge. It can be
used by users, databases and applications. The main
components of ontology are: concepts, relationships
and properties. Concepts represent “things” in a
given domain of interest, relationships connect these
concepts, and properties characterize concepts
attributes.
Existing and new applications can obtain benefits
from ontologies. For instance, different documents
assigned to the same ontology permit semantic
search that considers content instead keywords only.
Ontologies can also facilitate communication
between intelligent agents through vocabulary
sharing. As ontologies are usually expressed in a
logic-based language, tools can be used to reasoning
about concepts, relationships and properties,
inferring new concepts and offering automatic
support for intelligent services. Different ontology
languages and tools have been proposed. The OWL -
Web Ontology Language (Dean, 2004) is a W3C
recommendation that was used during SLA ontology
development.
3.1 OWL
The OWL language provides three sublanguages
designed for use by specific communities of
implementers and users. The first one is OWL-Lite
that supports classification hierarchy and simple
constraint features. The second one is OWL-DL,
which supports maximum expressiveness without
losing computational completeness. The third one is
OWL-Full that supports maximum expressiveness
with no computational guarantees.
An OWL ontology consists of Classes,
Individuals and Properties. Classes are a concrete
representation of concepts in domain we are
interested in. Individuals are instances of classes and
Properties are binary relations between Individuals.
There are two main types of properties: Object
properties and Datatype properties. Object properties
link an individual to an individual, e.g. hasSister
property. Datatype properties link an individual to
type, e.g. hasAge property.
OWL properties have specific characteristics,
which derive special properties. For example, each
object property may have a corresponding inverse
property, e.g. hasParent and hasChild properties.
Other property characteristics include functional,
inverse functional, transitive and symmetric. In
OWL, properties are used to create restrictions. As
the name suggests, restrictions are used to restrict
the individuals that belong to a class.
Restrictions in OWL fall into three main
categories: Quantifier Restrictions, Cardinality
Restrictions and hasValue Restrictions. There are
two quantifier restrictions: the existential (∃) that
can also be read as “someValuesFrom” in OWL
speak and universal quantifier (∀) or
“allValuesFrom” in OWL.
ADDING MEANING TO QOS NEGOTIATION
107

4 SLA ONTOLOGY
The QoS negotiation process usually involves
technical terms that are essential for providing
service but have no meaning for majority of web
users. Furthermore, the QoS contract (SLA)
commonly reflects this situation. The increase of
multimedia application, where users can perceive
quality, became clear the inadequacy of this model.
The QoS negotiation model proposal is based in
service personalization paradigm (Ribeiro, 2004b),
which considers the user QoS desired the start point
for service configuration. It also establishes a novel
format for QoS contract that explicitly includes
subjective quality requirements through which users
define QoS. Next sub-sections are dedicated to most
important elements of SLA ontology, namely classes
and properties. Section 5 depicts the role SLA
ontology plays in semantic QoS negotiation.
4.1 Classes
One of the most important elements of an ontology
is the concept. In OWL ontology, concepts are
described through classes. The concepts in SLA
ontology were extracted from ESCHER architecture
that is four layer QoS architecture for personalized
communication services based on quality of
perception (QoP) (Ribeiro, 2004b). According to
ESCHER, QoS negotiation depends on two other
processes: QoS specification that crosscuts different
perspectives (or layers) and QoS mapping that
involves these QoS information.
The ESCHER layers are: User, Application,
Middleware and QoS Mechanisms. The User Layer
is the most abstract perspective in which QoS
requirements can be defined. QoS information at this
layer represents the QoS level desired by the user,
represented by non functional requirements (NFR),
which are subjective and directly user perceivable,
e.g. “performance” and “cost”.
The Application Layer includes technical
requirement or QoS attribute, used for specialization
of subjective user QoS requirements. Applications
can be possibly categorized in real-time, interactive,
non-interactive and asynchronous classes. Examples
of QoS attributes are “frame rate” and “resolution”.
At Middleware Layer, the third layer, QoS
information is named QoS characteristic and it is
used for measurement purpose. Examples of QoS
characteristics are “delay” and “loss”.
Finally, the most concrete layer of ESCHER
architecture is the QoS Mechanism Layer. This is
actually an interface for QoS mechanisms that
effectively provide QoS guarantees, such as
Resource Reservation Protocol (RSVP). At this
layer, QoS information is defined as QoS parameter
used by QoS mechanisms. Figure 1 depicts these
QoS concepts, besides other ones that are used to
support QoS specification, such as VirtualResource,
used to abstract physical aspects.
Figure 1: SLA Ontology QoS Concepts.
In ESCHER, the QoScontract derived from QoS
negotiation and QoS mapping between QoS
specifications defined at different layers. The QoS
mapping process is helped by information about
QoS values (QoSMapTables), used during the
automatic mapping between user perspective
downwards QoS mechanisms, which actually realize
QoS guarantees.
The QoS contract is the most important concept
defined at SLA ontology. It serves as a basis for the
negotiation process and it explicitly describes the
QoS requirements stated by all participants of QoS
negotiation. As mentioned before, the QoS contract
proposal differs from traditional ones by explicit
inclusion of user perspective that make it a user-
friendly document.
Figure 2 details sub-concepts of QoS contract.
There are three components of QoS contract: (1)
UserQoSspec that includes QoS requirements at user
perspective; (2) ApplicationQoSspec that represents
QoS at application perspective; and (3)
ServiceQoSspec at service perspective.
The user QoS specification is made up by four
elements: (1) QoSdesired that defines high level user
requirements, like “High Quality of Audio” and
“Medium Cost”; (2) QoSperceived related to the
quality perceived by the user according personal
characteristics; (3) QoSpriorities that establish
priorities in case of impossibility to maintain the
QoS agreed, for example, the user can determine
that “Cost” is priority rather then “Performance”;
and (4) SatisfactionLevel that represents the opinion
of user about the quality actually provided and
perceived.
The Application QoS specification is dedicated to
QoS constraints imposed by the application. This
section is composed by three elements: (1)
ApplicationType that is used to make easier the
identification of the type of application by the user,
ICEIS 2006 - SOFTWARE AGENTS AND INTERNET COMPUTING
108

e.g. videoconference and Internet telephony; (2)
Category that identifies the application by the type
of QoS restriction: interactive, non-interactive, real-
time and Interactive real-time; and (3) QoSrequired
that represents a set of constraints imposed by the
type of the application.
Figure 2: SLA Ontology QoS Contract.
The service QoS specification represents the
technical section of SLA, similar to SLS – Service
Level Specification (Memenios, 2002). It is
composed by five elements: (1) QoSagreed that
represents the final product of QoS negotiation; (2)
QoSmeasured a set of pointers that makes possible
the evaluation of the QoS agreed; (3) QoSprovided
that is the actual QoS level provided by network; (4)
ServiceStatus that indicates the contract situation
(Active, Negotiating or Violated); and (5)
VirtualResourceAlloc represents the amount of
resource necessary to realize the service.
The resource negotiation process, as stated by the
ESCHER architecture, represents a mapping
between user QoS requirements into QoS parameters
used by QoS mechanisms, resulting resource
allocation. It is important to note that resources are
considered virtual resources. This strategy permits to
abstract complexity of physical aspects of real
resources without losing of expressiveness of QoS
specification.
4.2 Properties
There is a consensus about the existence of
relationships between QoS under user perspective
and QoS under provider perspective. However, the
nature of this relation and how it can help the
automation of personalized services is not clear.
Various properties were defined in SLA ontology to
address this aspect. Most of them are related to QoS
mapping process, i.e. they state relationships
between QoS information defined at different
perspectives.
Among properties related to QoS specifications,
there are six properties that are directly related to
non-functional requirements (NFR), listed at Figure
3. The first two are related to correlation, which is
defined in the scope of this work as an implicit
relation between distinct NFRs. The nature of
correlation can be: conflicting named
hasConflictingCorrelation, e.g. between “quality of
audio” and “usability”; or convergent, named
hasConvergentCorrelation, such as “performance”
and “cost”. This characterization concerns the
contribution of these requirements to user
satisfaction.
The treatment of correlation allows formalizing
knowledge about relationship between various
aspects that have influence over quality perceived by
end users. For instance, experimental results suggest
a relationship between performance and cost
(Steinmetz, 1995). When they are evaluated
together, each of these factors plays a negative
influence over the results of another. Figure 3
depicts these NFR related properties, declared as
existential constraints.
Figure 3: NFR related properties.
The other properties related to NFR are: (1)
isRealizedBy that relates subjective NFR to more
concrete QoS attribute; (2) isRelatedTo that relates
NFR to the application type; (3) isSpecializedBy that
relates a more abstract NFR to another, which
specializes it; and (4) the inverse property
(specializes).
There are two properties directly related to QoS
attributes. The first one relates QoS attribute to QoS
characteristic (producesImpactOn). It means that,
e.g. changes on QoS attribute “Frame Rate” produce
impacts over QoS characteristic “jitter” (delay
variation). The second property (realizes) is the
inverse of isRealizedBy property described early.
Figure 4 shows these properties.
Only one property that is directly associated to
QoS characteristic was defined in SLA ontology. It
ADDING MEANING TO QOS NEGOTIATION
109

is the inverse of producesImpactOn (see Figure 4),
namely isImpactedBy property.
Figure 4: QoS attributes related properties.
Figure 5 presents two properties associated with
the type of application (ApplicationType) class. The
first one assigns application type with the category
of application (Category), named isClassifiedAs
property. The second one named relates is the
inverse property of isRelatedTo property assigned to
NFR (see Figure 3).
Figure 5: Properties related to Application type.
A partial list of SLA ontology object properties
can be viewed through Figure 6. An emphasis is
made on isDirectlyUserPerceivable property that is
associated with user QoS specification. It means that
only a subset of NFR that is perceivable by user can
be used to specify QoS requirements at user level.
The bi-directional line relates property to its inverse.
One of the key features of ontologies that are
described using OWL-DL, as in this work, is that
they can be processed by a reasoner. A reasoner is a
tool that offers some services, such as to check
consistency of ontology classes and to automatically
compute inferred ontology class hierarchy.
5 SEMANTIC QOS
NEGOTIATION SCENARIO
This section deals with an implementation scenario
for semantic QoS negotiation. It details what
components are involved and how they can be used
for service discovery. The architecture proposal was
conceived as an extension of Service-oriented
architecture (SOA) (Erl, 2005). The original
components of SOA remain at extended architecture
(see Figure 7): the Service Requestor (or consumer),
the Service Provider and Service Repository (or
service registry).
Two new components were included at extended
SOA architecture proposal: Semantic Web Server
and ontologies. Besides the SLA ontology defined at
this work, two other ontologies are represented: Web
Services Ontologies (WS ontology), such as OWL-S
and WSMO, and specific ontology (or ontologies)
for NFR.
Figure 6: A partial list of SLA ontology object properties.
Semantic Web Services, including Semantic
QoS-aware ones, were summarized by the abstract
SWS component. At proposed architecture (see
Figure 7), mediators are represented by three
(possibly composed and complex) mediators (Feier,
2005): Negotiation Mediator concerning QoS Level
Mediation; Service Mediator that deals with Web
Service Mediation; and Ontology Mediator for QoS
ontologies merging or aligning. The concept of
mediation considered is generic, representing a
software agent that aims to solve conflict situations.
At architecture proposal, SOA activities were
modified in order to introduce QoS information. The
advertisement (publish) of service by the provider
(Service Provider), in a service registry (Service
Repository), was improved by including QoS level
offered (QoSoffered).
The client (Service Requestor) tries to discover a
service (find) at Service Repository, according to
user QoS specification (QoSdesired) and application
QoS requirements (QoSrequired). Discovery and
selection activities are made by semantic services
with mediation capacities and QoS information
(QoSdesired, QoSrequired and QoSoffered.
During execution phase (ServiceStatus Active),
the QoS level agreed can be monitored through
network mechanisms (QoSprovided), Middleware
services (QoSmeasured) or human physical senses
(QoSperceived). If degradation at QoS level is
detected (ServiceStatus Violated), the service can be
ICEIS 2006 - SOFTWARE AGENTS AND INTERNET COMPUTING
110

suspended and a renegotiation phase is started
(ServiceStatus Negotiating).
Figure 7: Architecture for Semantic QoS Negotiation.
6 CONCLUSIONS AND FUTURE
WORK
Traditionally, QoS negotiation is treated as a low-
level network mechanism. However, as the adoption
of service-oriented paradigm in application
development increases, dealing with QoS
negotiation between QoS-aware services has become
critical.
The use of ontology in the context of Web
Services is relatively a new research topic. OWL-S
and WSMO are important references in service
ontology field, but the way they describe services
are essentially by means of functionality, rather than
QoS capabilities. In this paper, an ontological
approach for QoS negotiation was detailed. It is
based on meaningful QoS contract defined through
SLA ontology.
Despite this work had been mainly focused on
end users and their participation during QoS
negotiation, it can be used by service-to-service
approach, i.e. without direct user interference. While
the offering of service-oriented application
increases, QoS becomes an important aspect to be
considered during application development, and then
it will be subject of further investigation. Using SLA
inside service composition is also subject of future
work, as well as the use of mediators to solve
potential conflicts and to merge and align
ontologies.
REFERENCES
Lawrence Chung, et.al. NonFunctional Requirements in
Software Engineering. Ed. Kluwer Academic
Publishers (2000).
David Booth. Web Services Description Language
(WSDL) Version 2.0 Part 0: Primer.
http://www.w3.org/TR/2005/WD-wsdl20-primer-
20050803/, (2005)
H. Peter Alesso and Craig F. Smith: Developing Semantic
Web Services. A K Peters LTD, (2004).
D. Martin, M. Burstein, O. Lassila, M. Paolucci, T. Payne,
and S. McIlraith. Describing Web Services using
OWL-S and WSDL. http://www.daml.org/services
/owl-s/1.0/owl-s-wsdl.html, (2003)
C. Feier (Ed.): WSMO Primer, WSMO Deliverable D3.1,
WSMO Working Draft, latest version available at
http://www.wsmo.org/2004/d3/d3.1/, (2005)
Chen Zhou et. al: Web Services Discovery with DAML-
QoS Ontology. Int. J. Web Service Res. 2(2): 43-66
(2005)
D.A. Menascé: QoS Issues in Web Services. IEEE Internet
Computing, vol. 6, no. 6, (2002)
C. M. F. A Ribeiro, N. S Rosa, P.R.F. Cunha. Meaningful
SLA: Towards a QoS Contract based on user
perception. Proc. 8th International IEEE EDOCS -
Workshop on Contract Architectures and Languages -
CoALa (2004)
FIPA. QoS ontology http://www.fipa.org/specs/fipa00094
/SC00094A.html, (2003)
J. Cardoso et.al: Modeling Quality of Service for
Workflows and Web Service Processes. Web
Semantics Journal, Elsevier,1(3): pp. 281-308 (2004)
H. Ludwig, A. Keller, A. Dan, R. P. King, and R. Franck.
Web Service Level Agreement (WSLA) Language
Specification, v1.0, (2003).
Mike Dean and Guus Schreiber, Editors, W3C
Recommendation, OWL – Web Ontology Language
Latest version available at http://www.w3.org/TR/owl-
ref/, (2004)
C. M. F. A Ribeiro, N. S Rosa, P.R.F. Cunha: Towards a
Model for Personalized Communication Services.
Proc. of 18th AINA, IEEE Press p. 99-102, (2004)
G. Memenios, G. Pavlou, D. Griffin, and L. Georgiadis.
Service Level Specification Semantics and
Parameters. Internet Draft, tequila-sls-02, (2002)
Ralf Steinmetz and Klara Nahrstedt: Multimedia:
computing, communications, and applications.
Prentice Hall, (1995).
Thomas Erl: Service-Oriented Architecture : Concepts,
Technology, and Design, Prentice Hall (2005)
ADDING MEANING TO QOS NEGOTIATION
111