Usages of Composition Search Tree in Web Service Composition

TL;DR

This paper explores the use of Composition Search Trees for efficient web service composition based on input-output matching, enabling the discovery of optimal service compositions like the shortest or leanest.

Contribution

It extends previous work by proposing the utility of Composition Search Trees to find specific types of service compositions such as shortest depth and leanest configurations.

Findings

Demonstrates the feasibility of using Composition Search Trees for web service composition.

Shows how different match types enable various composition strategies.

Provides a method to find optimal compositions like shortest depth and leanest configurations.

Abstract

The increasing availability of web services within an organization and on the Web demands for efficient search and composition mechanisms to find services satisfying user requirements. Often consumers may be unaware of exact service names that is fixed by service providers. Rather consumers being well aware of their requirements would like to search a service based on their commitments (inputs) and expectations (outputs). Based on this concept we have explored the feasibility of I/O based web service search and composition in our previous work. The classical definition of service composition, i.e., one-to-one and onto mapping between input and output sets of composing services, is extended to give rise to three types of service match: Exact, Super and Partial match. Based on matches of all three types, different kinds of compositions are defined: Exact, Super and Collaborative Composition. Process of composition, being a match between inputs and outputs of services, is hastened by making use of information on service dependency that is made available in repository as an one time preprocessed information obtained from services populating the registry. Adopting three schemes for matching for a desired service outputs, the possibility of having different kinds of compositions is demonstrated in form of a Composition Search Tree. As an extension to our previous work, in this paper, we propose the utility of Composition Search Tree for finding compositions of interest like leanest and the shortest depth compositions.