A CONTEXT-AWARE SEMANTIC WEB SERVICE EXECUTION
AGENT
António Luís Lopes and Luís Miguel Botelho
“We, the Body, and the Mind” Research Lab of ADETTI-ISCTE, Avenida das Forças Armadas
Edifício ISCTE, 1600-082, Lisboa, Portugal
Keywords: Service Execution, Semantic Web, Context-awareness.
Abstract: This paper presents the research on agent technology development for context-aware execution of semantic
web services, more specifically, the development of SEA (Service Execution Agent), a semantic web
services execution agent that uses context information to adapt the execution process to a specific situation,
thus improving its effectiveness and providing a faster and better service. Preliminary results show that the
introduction of context information and context-aware capabilities in a service execution environment can
speed up the execution process, in spite of the overhead that it is introduced by the agents’ communication
and processing of context information. The developed service execution agent uses standards such as
OWL-S service descriptions and WSDL grounding information. Also, an AgentGrounding definition has
been proposed to enable the execution of semantic web services provided by agents.
1 INTRODUCTION
Standard initiatives such as OWL-S and WSDL
(Martin, 2004) enable the automation of discovery,
composition and execution of semantic web
services, i.e. they create a Semantic Web, such that
computer programs or agents can implement an
open, reliable, large-scale interoperation of Web
Services. In this paper we describe part of our
research done on agent technology development for
context-aware execution of semantic web services.
We have decided to adopt the agent paradigm,
creating the SEA (Service Execution Agent) agent,
because we intend to integrate this work in open
societies of agents, enabling these to execute
semantic web services. (Paolucci, 2004) used the
same approach in the Web Services infrastructure
because of its capability to perform a range of
coordination activities and anonymising between
requesters and providers. In our research, the use of
context information helps improve the execution
process, by adding valuable situation-aware
information that will contribute to its effectiveness.
The major contributions of the present work to
advance the state-of-the-art are: i) the development
of a broker agent capable of executing received
OWL-S/WSDL service descriptions; ii) the inclusion
of context-awareness into semantic web services
execution. The rest of this paper is organized as
follows: section 2 gives a brief overview on the use
of context; section 3 describes the created broker
agent; section 4 presents some preliminary results of
the evaluation; section 5 concludes and presents
guidelines for future work.
2 CONTEXT-AWARENESS
Context-aware computing is a computing paradigm
in which applications can discover and take
advantage of context information to improve their
behaviour in terms of effectiveness as well as
performance. As described in (Dey and Abowd,
1999) context is any information that can be used to
characterize the situation of an entity. Entities may
be persons, places or objects considered relevant to
the interaction between a user and an application,
including users and applications themselves.
We can enhance this definition by stating that
applications are also considered to be entities. SEA
bases its activity in the context-aware computing
paradigm, where it uses and analyses context
information to enhance its service execution process,
by adapting it to the specific situation in which the
agent and its client are involved, at the time of the
execution process. This is done by interacting with a
231
Luís Lopes A. and Miguel Botelho L. (2007).
A CONTEXT-AWARE SEMANTIC WEB SERVICE EXECUTION AGENT.
In Proceedings of the Ninth International Conference on Enterprise Information Systems - SAIC, pages 231-234
DOI: 10.5220/0002376802310234
Copyright
c
 SciTePress

general purpose (i.e., domain independent) context
system (Costa and Botelho, 2005) for acquiring
context information, subscribing desired context
events and providing relevant context information.
Throughout the execution process, SEA provides
and acquires context information from and to this
context system. For example, SEA provides relevant
context information about itself, such as its queue of
service execution requests and the average time of
service execution. This will allow other entities in
the environment to determine the service execution
agent with the smallest work-load, and hence the one
that offers a faster execution service.
During the execution of a compound service,
SEA invokes atomic services from specific service
providers (both web services, and service provider
agents). SEA also provides valuable information
regarding these service providers’ availability and
average execution time. Other entities can use this
information to rate service providers or to simply
determine the best service provider to use in a
specific situation. Furthermore, SEA uses its own
context information (as well as information from
other sources and entities in the environment) to
adapt the execution process to a specific situation.
For instance, when selecting among several
providers of some service, SEA will choose the one
with better availability (with less history of being
offline) and lower average execution time.
In situations such as the one where service
providers are unavailable, it is faster to obtain the
context information from the context system (as long
as service providers can also provide context
information about their own availability) than by
simply trying to use the services and finding out that
they are unavailable after having waited for a
connection timeout to occur. If SEA learns that a
given service provider is not available it will contact
a service discovery agent or a service composition
agent requesting that a new service provider is
discovered or that the compound service is
re-planned. This situation-aware approach using
context information on-the-fly helps SEA to provide
a value-added execution service.
3 SEA: THE SERVICE
EXECUTION AGENT
The Service Execution Agent (SEA) is a broker
agent that provides context-aware execution of
services in the Semantic Web (whether they are
provided by web services or agents). The agent was
designed and developed considering the interactions
described in section 2. Its internal architecture was
clearly designed to enable the agent to receive
requests from client agents, acquire and provide
relevant context information, interacting with other
service coordination agents when necessary and
execute remote services.
3.1 Internal Architecture
The developed agent is composed of three
components: the Agent Interaction Component
(AIC), the Engine Component (EC) and the Service
Execution Component (SEC). Figure 1 illustrates the
internal architecture of the agent and the interactions
that occur between the components.
Figure 1: SEA Internal Architecture and Interactions.
The AIC was developed as an extension of the
JADE platform (Bellifemine, 1999) and its goal is to
provide an interaction framework to FIPA-compliant
agents (FIPA, 2002), such as SEA’s clients
(requesting the execution of specified services –
Figure 1, step 1) and service discovery and
composition agents (when SEA is requesting the
re-discovering and re-planning of specific services –
Figure 1, steps *). Among other things, the AIC is
responsible for receiving messages, parsing them
and processing them into a suitable format for the
EC to use it (Figure 1, step 2). The reverse process is
also the responsibility of AIC – receiving data from
the EC and processing it into the agents’ suitable
format to be sent as messages (Figure 1, step 9).
The EC is the main component of SEA as it
controls the agent’s overall activity by
pre-processing service execution requests,
interacting with the context system and deciding
when to interact with other agents (such as service
discovery and composition agents). When the EC
receives an OWL-S service execution request
(Figure 1, step 2), it acquires suitable context
ICEIS 2007 - International Conference on Enterprise Information Systems
232

information (regarding potential service providers –
Figure 1, step 3) and plans the execution process.
If the service providers of a certain atomic
service (invoked in the received composed service)
are not available, SEA interacts with a service
discovery agent (through the AIC – Figure 1,
steps *) to discover available providers for the
atomic services that are part of the OWL-S
compound service. If the service discovery agent
cannot find adequate service providers, the EC can
interact with a service composition agent asking it to
create an OWL-S compound service that produces
the same effects as the original service.
After having a service ready for execution, with
suitable context information, the EC sends it to the
SEC (Figure 1, step 4), for execution. Throughout
the execution process, the EC is also responsible for
providing context information to the context system,
whether it is its own information (such as service
execution requests’ queue, average time of
execution) or other entities’ relevant context
information (such as availability of providers and
average execution time of services).
The SEC was developed as an extension of the
OWL-S API (Sirin, 2004) and its goal is to execute
semantic web services (Figure 1, steps 5a and 6a)
described using OWL-S service description and
WSDL grounding information. The extension of the
OWL-S API allows for the evaluation of logical
expressions in conditioned constructs, such as the
If-then-Else and While constructs, and in the
service’s pre-conditions and effects. OWL-S API
was also extended in order to support the execution
of services that are grounded on service provider
agents (Figure 1, steps 5b, 6b). This extension is
called AgentGrounding and it is explained in detail
in (Lopes and Botelho, 2005). When the SEC
receives a service execution request from the EC, it
executes it according to the description of the
service’s process model. During the execution
process, SEC collects relevant context information
(such as providers’ availability, quality of service
and execution times). After execution of the
specified service and generation of its results, the
SEC sends them to the EC (Figure 1, step 7) for
further analysis and post-processing, which includes
sending gathered context information to the context
system (Figure 1, step 8) and sending the results to
the client agent (through the AIC – Figure 1, steps 9,
10).
3.2 Execution Process
OWL-S is an OWL-based ontology used to describe
semantic web services. OWL-S Services are
described in three parts: a Profile (which tells "what
the service does"); a Process Model (which tells
"how the service works"); and a Grounding (which
tells "how to access the service").
The general approach for the execution of
OWL-S services consists of the following sequence
of steps: i) validate the service’s pre-conditions,
whereas the execution process continues only if all
pre-conditions are true; ii) decompose the compound
service into individual atomic services, which in turn
are executed by evoking their corresponding service
providers (described in the grounding section of the
service); iii) validate the service’s described effects
by comparing them with the actual effects of the
service execution, whereas the execution process is
only valid if the service has produced the expected
effects; iv) collect the results, if any (the service may
be only a “change-the-world” kind of service), and
send them to the client who requested the execution.
4 EVALUATION
The enhancement provided by the use of context
information consists on the adaptation of the
execution agent to a specific situation (according to
available context information) in a way such that the
execution is done in the requested time-frame (by
the client agent) and that in case of failure of some
of the elements of the compound service, suitable
alternatives can be found. Using context
information, SEA can determine who the “best”
service providers are, by building a sort of
“reputation” schema of the available service
providers, which then can be used to determine the
fastest way to execute a compound service or find
alternatives in case of failure.
Even though this approach of service
coordination (the combination of service execution
with discovery and composition planning) improves
the way compound services are executed (by
allowing the determination of the best service
providers in a specific situation and by providing a
failure recovery method), it introduces an
“overhead” time that doesn’t exist in the actual
semantic web services execution environments. This
“overhead” time is composed of the procedures that
SEA must perform when communicating and
managing conversations with other agents (client,
service discovery and composition agents),
retrieving and processing context information
(related to the service providers) and preparing the
execution of compound services. It is important to
A CONTEXT-AWARE SEMANTIC WEB SERVICE EXECUTION AGENT
233

determine how this “overhead” time influences the
overall time of execution.
Figure 2: Execution test of a compound service with five
atomic services (280 sequential requests measured in
seconds).
Figure 2 shows the average behavior of SEA in
the sequential execution (280 times) of a compound
service (with 5 atomic services), by illustrating the
different parts of the execution process. The orange
area represents the Normal Execution Time (NET) –
this is the time that it takes to execute the web
services; the red area represents the Overhead
Execution Time (OET) – this is the time that SEA
takes to perform the mentioned procedures
(conversations’ management, acquiring context
information and preparing execution); the green area
represents the Total Execution Time (TET) – this is
the total time of execution: NET + OET.
The average “overhead” execution time is very
small (0,06 seconds). This value is hardly noticeable
by a user of such a system. Moreover, in some tests
the use of this kind of service coordination approach
allowed the reduction of the overall execution time
due to the fact that SEA always tried to find the
fastest service provider (the one with lower history
of “down” time, work load and average execution
time). This shows that SEA can be used in highly
dynamic environments by efficiently distributing the
“execution work” to the appropriate service
providers, hence providing a faster and more reliable
service to a user with time, device and location
constraints.
5 CONCLUSIONS
We have presented a framework to enable the
execution of semantic web services using a
context-aware broker agent. Test results show that
the introduction of context-aware capabilities and
the interaction within a service coordination
infra-structure not only add a very small overhead to
the execution process, as it turns SEA into a highly
efficient broker agent capable of distributing the
execution work among the available service
providers, thus providing a more useful and faster
service to its clients.
So far, SEA has been tested in an e-commerce
book search service environment and the results are
promising. The use of SEA in the CASCOM project
(Helin, 2005) will allow us to determine its
applicability and usability in different scenarios
operating in highly dynamic environments.
ACKNOWLEDGEMENTS
This work has been partly supported by the
European Commission under the project grant
FP6-IST-511632-CASCOM.
REFERENCES
Bellifemine F., Poggi A., Rimassa G., 2001. Developing
multi-agent systems with a FIPA-compliant agent
framework. Software-Practice and Experience 31 (2):
103–128 Feb 2001
Costa, P., Botelho, L., 2005. Generic Context Acquisition
and Management Framework. First European Young
Researchers Workshop on Service Oriented
Computing. Forthcoming.
Dey, A. K. and Abowd, G. D., 1999. Towards a better
understanding of context and context awareness. GVU
Technical Report GIT-GVU-99-22, College of
Computing, Georgia Institute of Technology.
FIPA Members. 2002. Foundation for Intelligent Physical
Agents website. http://www.fipa.org/.
Helin, H., Klusch, M., Lopes, A., Fernandez, A.,
Schumacher, M., Schuldt, H., Bergenti, F., and
Kinnunen, A., 2005. CASCOM: Context-Aware
Service Co-ordination in Mobile P2P Environments.
Multiagent System Technologies, Lecture Notes in
Computer Science, Vol. 3550 / 2005, ISSN: 0302-
9743, pp. 242 - 243
Lopes, A., Botelho, L.M., 2005. SEA: a Semantic Web
Services Context-aware Execution Agent. AAAI Fall
Symposium on Agents and the Semantic Web.
Arlington, VA, USA.
Martin, D., Burstein, M., Lassila, O., Paolucci, M., Payne,
T., McIlraith. S., 2004. Describing Web Services using
OWL-S and WSDL. DARPA Markup Language
Program.
Paolucci, M., Soudry, J., Srinivasan, N., Sycara, K., 2004.
A Broker for OWL-S Web Services. First International
Semantic Web Services Symposium, AAAI Spring
Symposium Series.
Sirin, E., (2004). OWL-S API project website.
http://www.mindswap.org/2004/owl-s/api/.
ICEIS 2007 - International Conference on Enterprise Information Systems
234