Parametrized automata simulation and application to service composition

This paper summarizes several results that have been published in [3]. Service Oriented Architectures (SOA) consider services as self-contained components that can be published, invoked over a network and combined with other services through standardized protocols in order to dynamically build complex applications. Service composition is required when none of the existing services can fulfill some client’s needs but a suitable coordination of them would satisfy the client requests. How to find the right combination and how to orchestrate this combination are among the key issues for service architecture development. Service composition has been studied in many works e.g. [9,4]. The related problem of system synthesis from libraries of reusable components has been thoroughly investigated too [8,5]. In this paper we address the composition synthesis problem for web services in which the agents are parametrized, i.e. the client and the available services exchange data ranging over an infinite domain and they are possibly subject to some data constraints. More precisely, the composition synthesis problem we consider can be stated as follows (e.g. [6,9]): given a client and a community of available services, compute a mediator, that can be viewed as a special service for enabling communication between the client and the available services in such a way that each client request is forwarded to an appropriate service. As usual (e.g. [4]), this problem is reduced to the computation of a simulation relation between the target service (specifying an expected service behaviour for satisfying the client requests) and the asynchronous product of the available services. If such a simulation relation is known then it can be easily used to generate a mediator, that is a function that selects at each step an available service for executing an action requested by the client. One of the most successful approaches to composition abstracts services as finite-state automata (FA) and apply available tools from automata theory to synthesize a new service satisfying the given client requests from an existing community of services. However it is not obvious whether the automata-based approach to service composition can still be applied with infinite alphabets since simulation often gets undecidable in extended models like Colombo ([1]). Starting from the approach initiated in [2] our objective is to define expressive classes