In this work we introduce global data computation in asynchronous distributed computation in comparison with the synchronous distributed computation. To increase the reliability of having all processes correct until the end of computation, and to increase the availability of having the global data vector correct and complete at the end of computation, a fault-tolerance-based system is proposed. In such a system processes are replicated m times. If m exceeds the maximum number of crashed processes, then the protocol terminates in one round. We prove that the reliability of a fault-tolerance-based system is better than that of a non-fault-tolerance-based system, we also prove that using higher values of m in a fault-tolerance-based system yields better reliability. The study reveals that a previously given protocol takes min(t+1, 2f+2) rounds, while the proposed protocol takes only one round.
展开▼
