Post-Failure Reconfiguration of CSP Programs

In this paper a technique called process merging is introduced. This technique allows the merging of two communicating sequential processes into a new single process. Thus, this technique can be used to reconfigure a distributed program after a faulty processing element has been detected. The technique is most applicable to dedicated multiple microprocessor systems where the need for continuous operation is critical. A process merging algorithm which operates on distributed programs using the CSP notation is presented in detail and its operation is discussed. In order to illustrate the merging technique, the algorithm's behavior is demonstrated using two classical distributed programs: the Bounded Buffer, Producer, Consumer program and the Dining Philosophers program. Finally, the merging technique is examined with respect to its demands on overall system operation and overhead. This examinatiQn leads to suggestions for future research.