Data Flow Computing Model: Application for Parallel Computer Systems Diagnosis

Abstract

This paper concerns the issue of recovery of a fault-tolerant parallel system without spares. A fault-tolerant system must be able to proceed its operation despite the occurrence of faults. In such a system recovery can be considered as a support function that retains the fault tolerant features of the system. We present here a data flow model that comprises functional blocks and activating signs in accordance with message flow in a parallel system. A fault in a system occurs according to the continuity of message routing providing the communication between processes and enables the fault diagnosis. Messages in a system are generated and processed in message routing.