On the Semantics of Coordination Models for Distributed Systems: The LogOp Case Study

Abstract

LogOp is a coordination model extending Linda by allowing a single coordination operation to dynamically address a multiplicity of possibly distributed tuple spaces. The design of LogOp raises relevant issues that are of general interest in the fields of coordination and distributed systems engineering. In particular, the design of an infrastructure that supports the enactment of coordination laws involving a multiplicity of physically distributed tuple spaces demands a careful treatment of the aspects related to synchrony, atomicity, locality of interactions, and global interpretation of coordination rules.In this paper we elaborate on these general issues, starting from the study of the semantics of LogOp. First of all, the LogOp coordination model is introduced as an extension of Linda. Then, two different semantics, both conforming to the LogOp informal specification, are formally described and compared. Finally, the limitations of the traditional approach to the formal characterisation of coordination (coordination as a language) are pointed out, and addressed through a different approach (coordination as a service) whose benefits are illustrated by suitably re-formulating LogOp semantics. On the one hand, this provides crucial hints on how the LogOp coordination model could be deployed as an interactive service provided by a coordination infrastructure for distributed systems. On the other hand, the above results allow some general aspects of coordination in open and distributed systems to be clearly pointed out and adequately discussed.