Talk @ ICWS 2010

1. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
SOA4All: An Innovative Integrated
Approach to Services Composition
Freddy Lécué1 Yosu Gorronogoitia2 Rafael Gonzalez2
Mateusz Radzimski2 Matteo Villa3
firstname.lastname@{manchester.ac.uk,
atosresearch.eu, txt.it}
1 The University of Manchester, Booth Street East, Manchester, UK
2 ATOS Research, Albarracín 25, Madrid, Spain
3 TXT e-Solutions SpA, Via Frigia 27, Milan, Italy
IEEE 8th International Conference on Web Services
July 5th - 10th , 2010
Miami, FL, USA

2. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Outline
1 Introduction
2 An Integrated Approach for Service Composition
3 Architecture Components
4 Experimental Results
5 Related Work
6 Conclusion

3. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Big Picture
(Semi-automation of) service composition in the semantic web.
Here we address ...
Semi-automated generation of optimal compositions wrt:
Properly formulated user requirements e.g., preferences,
constraints, goal;
On-demand generation of templates (or schema) based
composition.
So ...
Compositions are not elaborated from scratch!
... but rely on an end-to-end architecture of integrated
components: Template Generator, Composer and
Optimizer.

4. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Web service, Semantic Web and Semantic Web Services
Nowadays Web: syntax-based Web.
Semantic Web is an extension of current Web in which
information is given well-defined meaning.
Ontology: a key enabling technology (RDF, OWL)
Semantic web principles applied to web services
Give a semantics to services description;
Description languages with a semantics;
Bringing the web to its full potential SAWSDL, OWL-S,
WSMO ...
Web Services Intelligent Web
Dynamic Services
UDDI, WSDL, SOAP
WWW Semantic Web
URI, HTML, HTTP RDF, RDF(S), OWL
Static
Semantics 1

5. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
General Overview
Integration of 3 Main Components
Service Template Generator, Composer and Optimizer
Services Description
Relevant Relevant
Services Services
Repository of Compositions with
Services Discovery
Abstract Activities
Not Found
Compositions with
Design−Time Concrete Service Compositions
Template Composer Binding to Activities Optimizer
Generator
(ProM APIs) The end−user picks a (Parametric Design (Genetic Algorithms
template that fulfills based Approach) JGAP)
her goals
End User Constraints
and Preferences
Semantic
Reasoning
(Fact++) Execution
Execution Logs
(Active BPEL)
BPEL4SWS
Atomic Services

6. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
General Overview
Integration of 3 Main Components
Service Template Generator, Composer and Optimizer
Semantic−basedServices Description
(RDF Repository [PedLMLDK10])
Relevant Relevant
Services Services
Repository of Compositions with
Services Discovery ([JunAS10])
Abstract Activities
Not Found
Compositions with
Design−Time Concrete Service Compositions
Template Composer Binding to Activities Optimizer
Generator
(ProM APIs) The end−user picks a (Parametric Design (Genetic Algorithms
template that fulfills based Approach) JGAP)
her goals
End User Constraints
and Preferences
Semantic Semantic
Reasoning Links
(Fact++) ([LecL06])
Monitoring and
Execution
Execution Logs Management
Infrastructure
(Active BPEL)
BPEL4SWS
Atomic Services

7. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Integrated Approach: An Example
Context
Utilizing the cloud to expose (semantic web) services (with their non
functional properties) on product information (from different
companies) e.g., retrieving a product list, ordering a specific product.
Goal
Aggregation and classification of products from different service (e.g.,
getProductData, getProductURI, getProductPrice) providers.
ProductData ≡ ∀hasID.ProductID ∀hasName.ProductName ∀hasURI.URI
ProductURI ≡ ∀hasURI.URI
Product ≡ ∀hasID.ProductID ∀hasName.ProductName
∀hasURI.URI ∀hasPrice.ProductPrice
EuropeanProduct ≡ Product ∀hasProvider .Europe
ProductIDList ≡ ∀hasID.ProductID
ProductID , ProductPrice , Europe

8. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Template Generation (1)
Issue
Designing complex compositions: a difficult task for end-users.
Approach
Contextual-driven filtering
approach to discover relevant
past executions (logs) .
Process mining-oriented
techniques based on logs.
Semantic Reasoning on
services descriptions.
Discovery of possible templates
and their hierarchy.

9. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Template Generation (2)
Results
A selection of templates presented to end-users.
Advantages
Composition of activities/services: Not from scratch!
Collaborative generation of templates.
Customized configuration of templates, depending on the
abstraction level.
Services Discovery ([JunAS10])
Monitoring and Design−Time Compositions
Execution
Management Template Composer Optimizer
Infrastructure Generator (Parametric Design (Genetic Algorithms
(Active BPEL) (ProM APIs) based Approach) JGAP)

10. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Template Generation: An Example
Example
An E-Commerce template satisfying the goal: Aggregation and
classification of products of different service providers.
AbstractTask AbstractTask
GetProduct AggregateProduct
AND for Provider A
Branching by Provider A
AbstractTask AbstractTask
GetProduct ClassifyProduct
by Provider B by Provider B
Implicit Data Flow between Abstract Tasks of the Composition Task Start Node End Node

11. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Parametric Design Based Composition (1)
Issue
Goal-based, semi-assisted, and context-aware service composition.
Approach
Adaptable concretization
transformation using Multi-Agent
Systems e.g., adding new
ontologies, services and
templates.
Approach (Cont.)
Elaboration of services
binding and data flow from an
incomplete specification of a
composition.

12. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Parametric Design Based Composition (2)
Result
An executable composition of web services.
Advantages
Composition problem defined by domain-specific
sub-problems.
Adaptable and flexible version to achieve composition.
Knowledge-intensive configuration procedure.
Services Discovery ([JunAS10])
Monitoring and Design−Time Compositions
Execution
Management Template Composer Optimizer
Infrastructure Generator (Parametric Design (Genetic Algorithms
(Active BPEL) (ProM APIs) based Approach) JGAP)

13. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Parametric Design Based Composition: An Example
Example
Concretization of an E-Commerce template satisfying: Aggregation
and classification of products of different service providers.
AbstractTask AbstractTask
GetProduct AggregateProduct
AND for Provider A
Branching by Provider A
AbstractTask AbstractTask
GetProduct ClassifyProduct
by Provider B by Provider B
Implicit Data Flow between Abstract Tasks of the Composition Task/Service Start Node End Node
Semantic Link sl Input Parameter Output Parameter
getClothing aggregateClothing
ProductIDList ProductPriceData
ProductID getFootwear ProductPrice
ProductPrice
getFootwear European classifyEUFootwear
ProductIDList ProductIDList AND Branching Product ProductPriceData
ProductID
getFootwear ProductURI
ProductURI

14. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Optimization (1)
Issue
How select the most optimal among some achieving a same goal?
Approach
Criteria: Semantic and non functional properties of services.
Multi-objective optimization: Genetic algorithm-based approach.
Semantics based T2 T3 T6
Selection s3 s6
T1 s1, s2, s3, ...
2 2 2
T5 T8
s1 QoS based s5 s8
Selection
T4 T7
s4 s7

15. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Optimization (2)
Result
An optimal composition of services on complementary dimensions
i.e., semantic and non functional.
Advantages
A general and extensible model to evaluate compositions.
CSOP (Constraints Satisfaction Optimization Problem)
formalization.
Scalability of the Genetic algorithm-based approach despite the
off-line DL reasoning.
Services Discovery ([JunAS10])
Monitoring and Design−Time Compositions
Execution
Management Template Composer Optimizer
Infrastructure Generator (Parametric Design (Genetic Algorithms
(Active BPEL) (ProM APIs) based Approach) JGAP)

16. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Optimization: An Example
Example
Optimization of an E-Commerce composition satisfying: Aggregation
and classification of products of different service providers.
getClothing aggregateClothing
ProductIDList ProductPriceData
ProductID getFootwear ProductPrice
ProductPrice
getFootwear European classifyEUFootwear
ProductIDList ProductIDList AND Branching Product ProductPriceData
ProductID
getFootwear ProductURI
ProductURI
Semantic Link sl Input Parameter Output Parameter Service Start Node End Node
Service Updated by the Composition Optimizer
getClothing aggregateClothing
ProductIDList ProductPriceData
ProductID getFootwear ProductPrice
ProductPrice
getFootwear Product classifyFootwear
ProductIDList ProductIDList AND Branching ProductPriceData
ProductID
getFootwear ProductData
ProductData

17. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Scalability of the Architecture
Context
Up to 200 users u design compositions, up to 25 services s each:
(10u, 5s), (50u, 10s), (100u, 15s), (150u, 20s), (20u, 25s).
1e+09
Templates Generation
in Logarithm Scale
1e+08
Avg. Times (ms)
Composition
1e+07 From a Optimization
1e+06 Template
100000 of 2 act. 4 act. 7 act. 10 act. 12 act.
10000
1000
100
10
1
Case 1 Case 2 Case 3 Case 4 Case 5
10u;5s 50u;10s 100u;15s 150u;20s 200u;25s
Case i: Users u;Services s
Main Results
Template Generator: Number of logs, users and services.
Composer/Optimizer: Number of activities and services.

18. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Integration in the EU SOA4All project Architecture
Part of a SOA4All: Service-oriented Architecture for All
http://www.soa4all.eu/, http://coconut.tie.nl:8080/dashboard/

19. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
Related Work
Template Generator
R. Agrawal, D. Gunopulos, F. Leymann,
Mining process models from workflow logs
In EDBT, pages 469-483, ??, ??, 1998
Design Time Composer
P. Traverso and M. Pistore
Automated Composition of Semantic Web Services into Executable Processes
In ISWC, pages 380–394, Hiroshima, Japan, 2004.
Optimizer
L. Zeng, B. Benatallah, A.H.H. Ngu, M. Dumas, J. Kalagnanam, and H. Chang.
A Qos-aware middleware for web services composition
IEEE Trans. Software Eng., 30(5):311-327, 2004
As an End-to-End Integration
P. Bertoli and J. Hoffmann and F. Lécué and M. Pistore
Integrating Discovery and Automated Composition
In ICWS, pages 815-822, Salt Lake City, USA, July 2007.

20. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
An Integrated Approach to Services Composition
End-to-end composition and optimization:
Template Generator: Not to start from scratch.
Composer: Further details to the generated templates.
Optimizer: Maximization of the quality of compositions.
→ Validating in a realistic setting.
Future Work
Scalability issues: Template Generation:
Context-based filtering to improve relevance of logs.
Semantic-based mining for more relevant abstract process.
Adapting our architecture to contextual information.

21. Introduction Integrated Approach Architecture Components Experimentation Related Work Conclusion
An Integrated Approach to Services Composition
End-to-end composition and optimization:
Template Generator: Not to start from scratch.
Composer: Further details to the generated templates.
Optimizer: Maximization of the quality of compositions.
→ Validating in a realistic setting.
Future Work
Scalability issues: Template Generation:
Context-based filtering to improve relevance of logs.
Semantic-based mining for more relevant abstract process.
Adapting our architecture to contextual information.
Thanks for your attention!
Freddy Lécué freddy.lecue@manchester.ac.uk
http://www.soa4all.eu/, http://coconut.tie.nl:8080/dashboard/