DYNAMIC DISCOVERY OF WEB SERVICES USING MOBILE
AGENTS
Gaurav Tiwari, Rahul Agrawal, Shakti Mishra and Dharmender Singh Kushwaha
CSED, MNNIT, Allahabad, India
Keywords: Web services, Mobile agent, Web service discovery.
Abstract: Although UDDI has been extensively been promoted for discovery of web services, it fails to provide many
features like service provider validation, semantics lookup, Quality of Service Metadata, trust establishment
etc. We propose a mobile agent based approach to the discovery of web services. The mobile agent
framework is implemented using Java Agent Development Framework. The experimental results show that
the mobile agents reduce the need for bandwidth as only data transfer required is that of the mobile agent
itself. Since the mobile agents have all the data within themselves they need not communicate with other
system. This also reduces the load on the system and makes the system fault tolerant.
1 INTRODUCTION
Web services are typically application programming
interfaces (API) or web APIs that can be accessed
over a network, such as the Internet, and executed on
a remote system hosting the requested services Web
services provide the potential of implementing the
real world business functionality, because of its
flexible and adaptable framework. This flexibility
has allowed Web services to become part of many
existing application environment and has been one
of the reason behind their popularity (Erl, 2008).
Web services required an Internet-friendly and
Extensible Markup Language (XML) compliant
communication format that could establish a Simple
Object Access Protocol (SOAP) standardized
messaging framework.
Web services provide access to software systems
over the Internet using standard protocols. Web
Services Description Language (WSDL) is a
specification defining how to describe web services
in a common XML grammar (Cerami, 2002).
Microsoft, IBM and Ariba proposed UDDI
(Universal Description, Discovery, and Integration,
2000) to describe a standard for an online registry,
publishing and dynamic discovery of web-services
offered. Microsoft and IBM proposed WSDL as an
XML language to describe interfaces to web-
services registered with a UDDI database (Lu et al.,
2002).
Publishing a Web service involves creating the
software artifact and making it accessible to
potential consumers. Optionally a provider can
explicitly register a service with a Web services
registry such as UDDI or publish additional
documents intended to facilitate discovery such as
Web Services Inspection Language (WSIL)
documents. The service users or consumers need to
search Web services manually or automatically.
Discovery of web services can be carried out in
various ways and one of the way is to accomplish
this by using dynamic acting entities as mobile
agents. A mobile agent is a composition of computer
software and data which is able to migrate from one
computer to another autonomously and continue its
execution on the destination computer (Miraikar,
2006).
When a mobile agent decides to move, it saves
its own state, transports this saved state to the new
host, and resumes execution from the saved state.
Mobile agents are active entities that are free to
migrate between computers at any time during their
execution (Ketel, 2009). It makes these entities a
powerful tool for implementing distributed
applications in a computer network. Web-services
are registered with some public registries, intelligent
mobile agents might migrate from one service
registry to another to find desirable web-services as
specified by a user. Letizia (Lieberman) is a user
interface agent that assists a user browsing the
World Wide Web based on the user's profile.
497
Tiwari G., Agrawal R., Mishra S. and Singh Kushwaha D..
DYNAMIC DISCOVERY OF WEB SERVICES USING MOBILE AGENTS.
DOI: 10.5220/0003143604970500
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 497-500
ISBN: 978-989-8425-29-4
Copyright
c
 2010 SCITEPRESS (Science and Technology Publications, Lda.)

We have designed and implemented a framework
based on mobile agents for the discovery of web
services. A bunch of mobile agents migrate from one
machine to other collecting information about web
services and reporting them.
2 PROPOSED ARCHITECTURAL
FRAMEWORK
We propose a mobile agent based approach to the
discovery of web services. Our approach has three
major sections.
• The first deals with the discovery of web services
and reporting them.
• The second deals with publishing the web
services.
• Finally, the third deals with handling incoming
requests for finding the best possible web service
from all the available services.
2.1 Components of Architecture
Service Requester. It is the node that issues request
for a certain type of service according to its
requirements and it would include any or all of a list
of parameters like service name, type and number of
inputs, return type etc.
Request Optimizer.
It tries to deliver the most
similar web service matching the given requirement.
Service Broker
. It is the repository of all the web
services. There can be more than one service brokers
in the system.
Web Service Provider.
Web service providers
(WSP) actually publish the web services.
Web Service Repository.
The web service
repository is kept in the form of a relational
database. The following are the fields of the
database are Hostname, Service name, Function,
Input, Output, Install date, Discovery date and Report
date.
Controller Agent.
The controller agent is
responsible for the creation and management of other
mobile agents. It is also responsible for determining
the course of migration of the agents.
2.2 Functions of Mobile Agents
The mobile agent performs three functions:
1. First, it sends query to all machines for the list of
web services installed and compares this list with
the list that it already has discovered or the
changes that have been done to the old ones.
2. Second, when the agent reaches one of the
designated nodes, it reports the collected data
(list).
3. The agent would also collect performance data
about the nodes. This includes performance
metrics like network load, physical memory,
processor usage so that the service requests could
be matched with the most efficient machine on the
network.
The migration of the agent is managed by a controller
agent. The controller agent would act as the
scheduler for deciding the migration of agents.
2.2.1 Discovery & Reporting of Web
Services
Identifying Web Services. The mobile agent
searches the local system for the installed web
services & scans the interface where the WSDL of
the web services are published.
Collecting Performance Data. The agent also
collects performance Metrics. These performance
metrics would help the request optimizer to find the
best match for an incoming request.
Report the Web Services. When the agent arrives
at the designated node for the reporting of web
services, it updates the local database.
2.2.2 Publishing a New Web Service &
Updating Existing Web Service
The node deciding to publish a new web service
would first create its description using the standard
WSDL and develop the corresponding
implementation for the service, thereby publishing
the description.
2.2.3 Finding the Closest Match to the
Discovery Request
The service broker has the list of all the web services
with it. It matches it with the best possible service
that it has in its repository. If there are more than
one web services matching the request, the service
broker can reply back with a list of all of these
services to the requester. However, the broker can
break the tie on the basis of the performance criteria
collected by the agent.
KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development
498

2.3 Migration
The mobile agent uses the services of the Controller
Agent (CA) in order to decide the next machine to
migrate to. It calls the Controller Agents scheduling
algorithm with its current location. The controller
agents decide the next location of the calling agent
(fig. 1).
Figure 1: Migration of Mobile Agents.
Round Robin Migration. Under this scheme the
agent sequentially migrates through each and every
node in the system one by one (Fig.2).
Figure 2: Round Robin Algorithm for Migration of Mobile
Agents.
Most Active Next Migration. This migration
algorithm (Fig.3) visits the most active nodes more
than the less active node.
The controller agent
maintains a count of the web services created and/ or
modified by all the machines. To select the next
destination, the controller agent uses a priority
queue. To avoid starvation, relatively higher weight
should be given to time since last visit.
3 TOOLS AND FRAMEWORKS
USED
We have used Apache Axis, which is an open source,
XML based Web service framework. It consists of a
Java and a C++ implementation of the SOAP server,
and various utilities and APIs for generating and
deploying Web service applications. The mobile
Figure 3: Most active Next Migration Algorithm.
agent framework is implemented using JADE (Java
Agent Development Framework) fully implemented
in Java language.
3.1 Creation of JAVA Web Service
In our proposed system, a Web Service is a JAVA
Class. The public functions of the class would be
available when the service is published.
3.2 Deployment of Web Service
Once the WSDL and Web Service are ready, we
create a deployment descriptor. The deployment
descriptor describes the method to deploy the web
service. In the deployment descriptor, we allow or
disallow routines in the web service.
4 EVALUATIONS
For each set of values, we have calculated the
average delay. To simulate different sequences of
events such as installation, discovery time and
reporting time, we take different number of nodes to
manage varying number of agents.
4.1 Computing Effect of Number of
Agents on the Average Delay
between Installation and Discovery
Time of Web Services
We observe from fig. 4 that as the number of
machines increase, the candidates for inspection by
the mobile agent increases. We can see that the
increase in number of machines causes the average
Global: priorityArray[ ].
Initialize priority Array with all having zero
priority.
MOST ACTIVE NEXT(currentid,locationlist)
{
//update priorityArray
For each location in location list
Find the percentage of recent change
performed by it
Calculate the time since any agent had
last visited the node
update the priority in the priorityArray
destinationID=index with maximum priority
return destination ID
}
ROUNDROBIN-next-destination
(currentnodeID,locationlist)
{
next-destinationID = curentnodeID+1;
next-destinationID%=sizeof(locationlist)
return next-destinationID
}
DYNAMIC DISCOVERY OF WEB SERVICES USING MOBILE AGENTS
499

delay between installation and discovery time of the
web services to increase. As the number of agents
increase, the workload for one agent decreases. As
the agents work asynchronously and parallel, the web
services are discovered in shorter span of time.
Figure 4: Effect on number of agents on the average delay
between installation and discovery of web services.
4.2 Average Delay between Installation
and Reporting Time of Web
Services
We observe from fig. 5 that as the number of
machines where the data has to be reported increase,
the agent would have to stop at more location to
submit the data. This causes an increase in the
average delay between installation and report time.
Figure 5: Effect on number of agents on the average delay
between installation and reporting of web services.
4.3 Average Delay in Installation and
Discovery
We observed from fig. 6 that as the number of web
services increases, the agent requires more time to
check for the new services or the changes in the old
ones. Thus, the average delay increases as there are
more services to be checked and reported.
Figure 6: Effect on total number of web services on the
average delay in installation and discovery.
5 CONCLUSIONS
This paper presents the mobile agent based approach
to the discovery of web services. Although UDDI has
been extensively been promoted for discovery of web
services, it fails to provide many features like service
provider validation, semantics lookup, Quality of
Service Metadata, trust establishment etc. We have
shown that the mobile agents reduce the need for
bandwidth as only data transfer required is that of the
mobile agent itself. We have also shown that the
mobile agent based system is fault tolerant. As there
are more than one mobile agent in the system, even if
some fail the others could complete the task and
would not allow the system to go down. Thus the use
of a mobile agent based approach is a good
alternative to the conventional approach using
centralized servers to report web services.
REFERENCES
Bellifemine, F, et Al. JADE A White Paper.
Pautasso, C., Limitations of SOAP, WSDL, UDDI.
Computer Science Department, Swiss Federal Institute
of Technology.
Gengler, B., 2001. UDDI Has Problems . Computer Fraud
& Security. Volume 2001, Issue 8
Chapell, D., article_who_cares_uddi.html.
http:// www.davidchappell.com/articles/
Lu, S., Dong M. and Fotouhi, F., 2002.The Semantic Web
opportunities and challenges for next-generation, Web
applications ”, Information Research, Vol. 7 (4)
Lieberman, H. Letizi An Agent That Assists Web
Browsing.
Erl, T., 2008. Service-Oriented Architecture. Pearson
Education
Cerami, E. , 2002. Web Services Essentials, O’Reilly
Miraikar, Z., 2006. Resource and Service Discovery for
Mobile Agents Platform.
Ketel, M. 2009. A Mobile agent Based Framework for
Web Services”, IEEE Internet Computing V. 10(3), pp.
58-65.
KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development
500