Consistency of Replicated Log for Service Recovery

The requirements in terms of reliability and availability have become crucial in the context of many modern information systems. Therefore, fault-tolerance support in these systems becomes pivotal. Among possible techniques used to increase fault-tolerance are checkpointing and message logging. Both, checkpoint and message log, are commonly said to be saved in a stable storage, which is an abstraction of a perfect storage, that survives failures, and thus ensures the stability of recovery data. In this paper, we propose to implement a stable storage abstraction by replicating logs that contain recovery information. For this purpose, we introduce the eager request replication protocol, and discuss its functioning in case a failure of any system component occurs. The proposed solution is especially tailored for the service oriented systems, though it is not limited to them.