Service Composition Issues for Distributed Business Processes

Abstract

One definition of the "on demand era" is that customers will be able to obtain services and goods as and when they need them. Many customers will need a composition of several services (a portfolio) to execute their business processes. The choice of service providers for business process activities and the certainty of their combined commitment determine the efficiency of the process. The selection of service providers is difficult because of the complex relationships among the process activities in cost, time, distance, etc.; selecting one flight from LA to London can impact the choice of the return flight from London. To ensure the efficiency of the whole process the portfolio of chosen services must be acquired atomically; either all services are acquired or none of them are. This paper deals with algorithms and infrastructures that allow customers to select and obtain optimal service portfolios for business processes. Specifically, we address two problems we believe are central to service composition: How to select optimal component services so as to maximize the utility of a service portfolio subject to activity relationships, and how to reserve all the services in a portfolio atomically.