RSSM: A Rough Sets based Service Matchmaking Algorithm

1
RSSM A Rough Sets based Service Matchmaking
Algorithm

• Bin Yu and Maozhen Li
• School of Engineering and Design
• Brunel University
• Uxbridge, West London

2
Outline

• Service Oriented Architecture
• Service Oriented Grids
• Web/Grid Service Discovery Methods
• RSSM Service Matchmaking with Rough Sets
• Problems RSSM tries to solve
• How it works
• Preliminary results achieved

3
Service Oriented Architecture (SOA)

• SOA is emerging as the mainstream architecture
  for developing next generation interoperable Grid
  systems
• Facilitated by OGSA, a standard architecture for
  developing service oriented Grid systems
• Enabled by Web services technologies such as
  WSDL, SOAP, WSRF, WS-N and many others.
• Engineered with technologies such as Apache Axis,
  GT4, WSRF.NET
• SOA define standard interfaces and protocols that
  allow developers to encapsulate resources
  (computers, data, storage, software, instruments)
  as services
• Benefits of SOA
• Makes future heterogeneous Grid systems
  interoperable using open standard technologies
• Facilitates service discovery for use by a wider
  user community

4
Service Oriented Grids

• What is a service?
• A service is a software component that
• Has a rich describable interface (e.g. WSDL)
• Can be advertised, discovered, composed, located
  via computer programs not by human beings
• Has operations on underlying resources
• Another Grid definition in the context of SOA
• A Grid is a large scale distributed system that
  couples a bunch of services utilising various
  resources on the Internet

5
Service Discovery

• UDDI
• An industrial standard for Web services
  registration and discovery
• Major Limitation Only supports keyword match
• Service discovery with Semantic Web technologies
• UDDI-M, OWL-S, UDDI/OWL-S, WSMO
• Using metadata to annotate service entities
  (properties), and using ontologies to describe
  the relationships between metadata
• Relationships of service properties include
  exact, plugin, subsume, nomatch.

6
Introducing RSSM

• Service Matchmaking with Rough Sets
• Rough sets is a mathematic tool that can be used
  to find knowledge from incomplete data sets
• Problems RSSM tries to solve
• Be able to deal with uncertainty of service
  properties when matching service advertisements
  with service requests
• This issue has never been explored before
• Why this issue exists?
• Because of the scale of Grid, service providers
  may advertise services independently
• using independent properties to describe services
• and same to service requesters

7
How RSSM Works
Advertised Domain Services
Domain Ontology
3
4
7
2
Irrelevant Property Reduction
A Service Query
1
9
5
6
Matched Services
Dependent Property Reduction
Service Matching and Ranking
8
10
8
Irrelevant Service Property Reduction
9
Dependent Service Property Reduction

• By dynamically reducing irrelevant and dependent
  properties in terms of a service request, RSSM
  can deal with uncertain properties in service
  discovery.

10
Service Matchmaking and Ranking

• Properties of service advertisements and service
  requests are matched based on their relationships
  from 100 to 60
• Properties without values (uncertain properties)
  have a match degree of 50
• Calculating a match degree using

11
RSSM Implementation

• Java
• mySQL as a service repository
• It now provides interfaces to upload and parse
  OWL-S files for service registration
  automatically using OWL-S API
• It also provide an interface to register services
  in a UDDI in the service repository

12
RSSM Effectiveness

• We compared RSSM with UDDI, and OWL-S Classical
  Matchmaking Method Proposed by M. Paolucci et al
  (at ISWC 2002)
• Used a Pentium IIII 2.6G with 512M RAM running
  Red Hat Fedora Linux 3
• Used jUUDI and mySQL to build a UDDI registry
• Extended Pizza Ontology
• http//www.co-ode.org/ontologies/pizza/pizza_20041
  007.owl
• Used Protégé OWL API to Parse OWL Documents
• Used RACER as an inference engine
• Results

13
RSSM Overhead
Overhead in retrieving service records
Overhead in matching service records
14
Concluding Remarks

• Service discovery is of paramount importance for
  a wider uptake of the Grid
• The preliminary results of RSSM are encouraging
• But reducing dependent service properties is
  computationally intensive, which needs some
  optimisation work