Application controlled checkpointing coordination for fault-tolerant distributed computing systems

摘要

In order to provide fault tolerance for distributed systems, the checkpointing technique has widely been used and many researches have been performed to reduce the overhead of checkpointing coordination. In this paper, we present a new checkpointing coordination scheme in which the application controls the coordination activity by utilizing the commu- nication pattern of the application program. Unlike the previous solutions which do not utilize the communication pattern of cooperating processes. it is possible to reduce the coordination effort as well as the number of checkpoints enforced to be taken. Extensive simulations have been performed to evaluate the proposed scheme and we have concluded that the proposed scheme significantly reduces the coordination overhead compared with the existing loose coordination scheme.