Satya Sanket Sahoo

1
Web Services Composition(An AI-based Semantic
Approach)

• Research Seminar (CSCI 8990)

Satya Sanket Sahoo Department of Computer
Science University of Georgia
2
Outline of my Talk

• Web Services
• Web Services Composition
• Approaches to WS Compositions
• Semantic Web Approach
• DAML-S
• Situational Calculus
• Petri Nets
• Discussion AI based approach to WS Composition

3
Service Oriented Computing
Background

• New paradigm to build complex web-based
  applications
• Platform-independent software components
• User-defined, need-based composable
• Web Services will
• Enable users to access business functionalities
• Enable streamlining and enhancement
  heterogeneous enterprise application integration
  

Web Services
4
Web Service Framework
Background
Service Registry
Find
Publish
Service Consumer
Service Provider
communicate
5
Problem

• To achieve a given objective sell my car, I may
  need more than one Web Service
• Car Valuation service WS1
• Car Auction Service WS2
• Online Financial service WS3
• All three WS need to interact seamlessly

To achieve my goal
6
Web Services Composition

• Composed Web Services are
• Individual components implemented at different
  places
• Execute in different contexts
• But, need to communicate to yield desired
  behavior
• Multiple approaches based on perspective
• Static Dynamic WS composition
• Industry solution Semantic Web solution

7
Static Dynamic WS Compositions
Static Dynamic
Web Service Composition
Static Composition
Dynamic Composition
By Hand/ hardcode
BPEL4WS
Model driven Service composition
XSRL
8
Industry Solution
Industry Semantic Web

• Interface design of a web service is in WSDL.
• Interaction and message exchange between
  different WS BPEL4WS or WSCL
• Composing a process using the above languages
  involves specifying
• Role of each participating Web Service
• Logical flow of messages between them

9
Semantic Web
Industry Semantic Web

• WS annotated regarding their functionalities,
  domain, location.
• Annotations are concepts enable software
  agents to understand usage of a WS
• concepts are defined in an ontology
  standardize their interpretation
• Ex OWL-S ( an OWL ontology for specification of
  WS), DAML-S

10
DAML-S
DAML-S Web Service Description
Service Class
properties
presents
describedBy
supports
Classes as ranges
Service Profile
Service Model
Service Grounding
11
Service Profile
DAML-S Web Service Description
Gives high level description of WS
Input, output, preconditions, postconditions
Used to advertise its features
Allows clients to select and locate WS from
registries
12
Service Model
DAML-S Web Service Description
Detailed description of service
Process model
Process control model
Service Grounding

• Binding level information
• specifies how client can access the service
  (SOAP/Java RMI)

13
Coming back to WS composition
WS2 (Auction Service)
Composed of
Buyer Interface service
Seller Interface service
Financial authentication service
May be executing in parallel to different
instances of a transaction
14
Semantic Approach to WS composition

• Apply logical inferencing techniques to compose WS

WS capability Annotated with DAML-S
Situational calculus
Translate into a More expressive logic
Situational Calculus
DAML-S
SC precondition
Precondition
SC effects
Effects
SC knowledge precondition
Inputs
SC knowledge effects
Output
15
Situational Calculus

• First order logical language represents
  dynamically changing worlds
• All changes direct result of named actions

Action a(y)
Situation (s)
Situation (So)
Results in
16
Situational Calculus to Petri Nets

• The representation in SC translated into Petri
  nets
• Petri nets bipartite graphs

Places
Transitions
Places
Places
Input place
Output place
17
Capabilities of Petri nets

• Formal definition a Petri net (PN) is an
  algebraic structure (P,T,I,O) composed of
• Finite set of places, P p1, p2, , pn
• Finite set of transitions, T t1, t2, , tm
• Transition input function, I I maps each
  transition ti to a multiset of P
• Transition output function, O O maps each
  transition ti to a multiset of P
• Allows to model events and states in a
  distributed system
• Cleanly captures sequentiality, concurrency and
  event based asynchronous control

18
DAML-S composite process Petri net
Start
Finish
Component Control Construct
Ready
Done

• Distributed operational semantics
• each transition fires based on local input
  condition
• Transition firing correspond to system evolution

19
So, the big picture
DAML-S Annotated WS
Situational Calculus representation
Petri Nets representation
Composed Process
User-template
DAML-S Annotated WS
Reasoner Ex GOLOG reasoner

• SC based reasoner given user defined template
  runtime binding to goal instance
• Reasoner evaluates non-deterministic choices and
  executes the plan

20
Discussion

• Problems with AI-related approach
• Complex control structures like loops,
  nondeterminism and choice cannot be used
• Unlike classical planning, all objects are not
  available in initial state new objects may be
  created at runtime.
• Planning deals with sparse objects, though web
  services use typed messages rich structure

21
Conclusion

• There are several approaches to compose multiple
  web services single business process
• Both, BPEL4WS and SC based program are manually
  written
• No assembling of complex flows from atomic
  message exchange search process based

22
References

• Simulation, Verification and Automated
  Composition of Web Services Srini Narayanan,
  Sheila McIlraith
• Web Services Composition Current Solutions and
  Open Problems Biplav Srivastava, Jana Koehler
• Augmenting Semantic Web Service Description With
  Compositional Specification Monica Solanki,
  Antonio Cau, Hussein Zedan
• Approaches to Web Service Composition Dimitry
  Shaparau
• Semantic e-workflow composition J. Cardoso, A.
  Sheth

23
References

• Planning with Workflows An Emerging Paradigm
  for Web Service Composition Biplav Srivastava,
  Jana Koehler