An E cient Logging and Recovery Scheme for LazyRelease Consistent Distributed Shared

Checkpointing and logging are widely used techniques to provide fault tolerance for the distributed systems. However, logging imposes too much overhead on the processing to be a practi-val solution. In this paper, we propose a low-overhead logging scheme for the distributed shared memory system based on the lazy release consistency model. Unlike the previous schemes in which the logging is performed when a new data item is accessed by a process, the logging in the proposed scheme is performed only when a process performs a release operation following any writes, and hence the logging frequency can be signiicantly reduced. Moreover, instead of logging the data items accessed at each process into the stable storage, a process logs the data items it has written into the volatile log and only the vector time and other access information are logged into the stable storage. During the recomputation after a failure, the correct version of data items accessed at each process can be eeectively traced by using the logged vector time. As a result, the amount of logging at each process can also be reduced.