Service Discovery Using Communication Fingerprints

Abstract

A request to a service registry must be answered with a service that fits in several regards, including semantic compatibility, non-functional compatibility, and interface compatibility. In the case of stateful services, there is the additional need to check behavioral (i.e. protocol) compatibility. This paper is concerned with the latter aspect. An apparent approach to establishing behavioral compatibility would be to apply the well-known technology of model checking to a composition of the provided service and the requesting service. However, this procedure must potentially be repeated for all provided services in the registry which may unprohibitively slow down the response time of the broker. Hence, we propose to insert a preprocessing step. It consists of computing an abstraction of the behavior for each published service that we call communication fingerprint. Upon request, we use the fingerprint to rule out as many as possible incompatible services thus reducing the number of candidates that need to be model checked for behavioral compatibility. The technique is based on linear programming and is thus extremely efficient. We validate our approach on a large set of services that we cut out of real world business processes.