An Efficient Fault-Tolerant method for Distributed Computation Systems

Fault tolerance is one of the most important features required by many distributed systems. We consider the efficiency issues of constructing distributed computing systems that can tolerate Byzantine faults. The well-recognized technique is to introduce replicated computation and derive the correct results through a voting mechanism. While this technique is applied to each computation request individually, we believe that by considering multiple requests at the same time in a distributed environment, we can greatly improve its efficiency. This is based on the observations that computation requests may be ordered in a different way for computation at different nodes, and the verdict of the correct result for one request may imply the correct result for another request. We propose to exploit a suitable solution to improve the efficiency of the existing technique to avoid unnecessary computation and unnecessary message exchanges among distributed processes