Model and Algorithm of Backward Error Recovery of Distributed Software

Backward error recovery is one of the important techniques of software fault tolerance. Because of error propagation its recovery in distributed software needs cooperation between processes to achieve consistent recovery. However, the techniques of the achievement suffer from either concurrency level decreasing or the domino effect. Based on a formal model of the distributed system, a backward recovery protocol without the two drawbacks is specified in this paper. The algorithm of the protocol is proven strictly and its implementation is proposed.