New Causal Message Logging Protocol with Asynchronous Checkpointing for Distributed Systems

Abstract. Causal message logging is an efficient approach for tolerating failures of processes in distributed systems because it has the advantages of both pessimistic and optimistic message logging approach. However, traditional causal message logging protocols prevent live processes from executing continuously their computation and require some synchronous logging to the stable storage during recovery. Although Elnozahy’s protocol solves the problems, it has the central recovery leader’s problem. Additionally, if it were integrated with asynchronous checkpointing, it may result in inconsistency problems in case of concurrent failures. In this paper, we present a new causal message logging protocol with asynchronous checkpointing to need to maintain only the latest checkpoint of each process and allow live processes to execute continuously their computation even in concurrent failures during recovery. Moreover, the protocol solves the problems of Elnozahy’s protocol and improves asynchrony during recovery because the protocol enables each recovering process to be responsible for only its recovery. 1