A High-Performance Recovery Algorithm for Distributed Systems

摘要

In this paper, a high performance recovery algorithm for distributed systems is presented. It offers fast determination of the recent consistent global checkpoint (maximum consistent state) of a distributed system after the system recovers from a failure. We have assumed that the system uses asynchronous check pointing scheme for taking checkpoints. The main feature of the proposed recovery algorithm is that it is executed simultaneously by all participating processes and it avoids unnecessary comparisons of checkpoints while determining a consistent global checkpoint of the system. It ensures fast execution. We have also presented an enhancement of the proposed recovery algorithm to achieve even a simpler recovery.