Web Service Discovery Mechanisms
Looking for a Needle in a Haystack?
Evangelos Sakkopoulos
joint work with
J. Garofalakis, Y. Panagis, A. Tsakalidis
University of Patras, CEID
& RA Computer Techonology Institute
Overview of Talk
Introduction
 Description of Players
 Discovery Architectures
 Data Models
 Quality of Service
 Conclusions

2
Research Academic Computer
Technology Institute
The name of the game


Web Service (WS): interoperable S/W
components that can be used in
application integration and component
based application development
WS Discovery:
requester + middle agent = find

3
Find the WS matching certain functional
criteria
Research Academic Computer
Technology Institute
Reasons for discovery +
common problems
Why:
 Need for complex WS invocation
patterns
 Need to chose between several
descriptions
Problems:
 Heterogeneities in Technical,
Pragmatical, Ontological Level
4
Research Academic Computer
Technology Institute
Description of Players
[catalogues]


Centralized repositories of WS
Descriptions
UDDI – emerging protocol, v. 3.0
3.
SOAP APIs
XML representation for the registry
WSDL interface definitions
4.
APIs Defs of various tech. models
1.
2.

Three types of info available in UDDI



5
White pages (contact info)
Yellow pages (WS categorization)
Green pages (Technological INformation)
Research Academic Computer
Technology Institute
Description of Players
[P2P systems]





6
Distributed, Load balanced repositories
Typical P2P overlay Chord [Stoica et. al. 2001],
Pastry [Rowstrom et. Al. 2001], CAN
[Ratnasamy et. al. 2001]
Several WS Discovery systems have chosen
Chord as overlay
WS Descriptions hashed and distributed over
Chord Ring
Speed-R [Sivashanmugam et. al. 2004], uses
combination of Ontological mapping and P2P
(nodes have different roles and are controlled
by ontologies)
Research Academic Computer
Technology Institute
Discovery Architectures (1)

Manual


Automatic


UDDI registry: Centralized, authoritative
repository of service descriptions
Decentralized



7
Discovery by a requester agent
Centralized


A human queries and decides
Distant ancestors of Whois++, rWhois systems
UDDI Federations
P2P systems
Research Academic Computer
Technology Institute
Discovery Achitectures(2)

Following the standards:
Info is added on white or yellow pages
 Modify green pages (design by contract)


Ignoring the standards
Active UDDI (a new WS for mediator)
 Grid Computing


Industrial Standards

8
J2EE, MS .NET, Java-based APIs
Research Academic Computer
Technology Institute
Discovery Achitectures (3)
9
Research Academic Computer
Technology Institute
Data Models - The IR
viewpoint
UDDI keyword matching = Boolean
IR Model
 Sajjanhar et. al., 2003:

Service Descriptions are modelled as
texts, texts as vectors, a termdocument matrix A is built
 LSI is applied to A
 Ability to query by similarity

10
Research Academic Computer
Technology Institute
Data Models - The IR
viewpoint (cont’d)

Schmidt and Parashar, 2004
WS Descriptions = d-dimensional point
 Hilbert curve: Points are mapped to 1-d
and assigned unique IDs
 IDs hashed and distributed in a Chord


XChord, Li et. al.
P2P discovery
 Descriptions extracted, hashed and
distributed across Chord

11
Research Academic Computer
Technology Institute
Data Models- The
Semantics Viewpoint



Desideratum: retrieve WS with similar
functionality
Semantic WS descriptions with DAML-S OWL-S
Paolucci et. al., ISWC 2002





Sivashanmugam et. al., ISWC 2003

12
An ontology for each WS (Service Profile)
Service Profile: Functional Attr. , Functional Descr.
Ontology  Subsumption and Semantic Matching
Matchmaker implemented as UDDI add-on
Matching engine implemented with semantic additions to
WSDL descriptions
Research Academic Computer
Technology Institute
Data Models- The
Semantics Viewpoint (2)

Moreau et. al. 2002,



Hu, NODe 2002,



Domain Ontologies and Operation Ontologies
Binding Ontology performs matching
Overhage, 2002

13
Agents are described as WS
Matching: structural validations of queries against XML
service descriptions
Blue pages, a new UDDI section with semantic
descriptions
Research Academic Computer
Technology Institute
Data Models – A taxonomy
14
Research Academic Computer
Technology Institute
Quality of Service Concerns



Quality of Web Service (QoWS): a rather
neglected issue
First attempt to define: Ran, 2003
QoWS parameters:



15
Computational Behavior: Latency, Accuracy,
Throughput, Availability
Business Behavior: Invocation Cost, Company
Reliability
Metadata Constraints: Location, Company
Preference etc.
Research Academic Computer
Technology Institute
Discovery with QoWS

Ouzzani and Bouguettaya, IEEE Internet
Computing, March 2004.




16
QoWS parameters are categorized as negative
and positive.
QoS distance, measures advertised vs provided
QoWS
Execution plan, an ordered execution sequence
of WSs.
Selection performed of an optimum execution
plan that maximizes provided QoWS
Research Academic Computer
Technology Institute
Conclusions





17
Surveyed work emphasizes binding and
matching.
UDDI and P2P systems, the main players
Data models: classical IR to Ontologies
More emphasis to QoWS provisioning
Discovery not only for WS, web-based S/W
components, too.
Research Academic Computer
Technology Institute
Thank you
Inquiries:
[email protected]