Verification of JavaSpaces/spl trade/ parallel programs

Abstract

We illustrate a formal verification method for distributed JavaSpaces applications by analyzing a nontrivial fault tolerant algorithm that solves a typical coordination problem. The problem consists of the computation of an extensive task, performed in parallel by splitting it into smaller and more manageable parts. The proposed solution, based on JavaSpaces coordination primitives, transactions and timeouts, is verified by translating it to the formal language /spl mu/CRL, together with the previously developed /spl mu/CRL-model of the JavaSpaces architecture, and by using model checking techniques.