A hybrid roll-forward recovery scheme for distributed systems

摘要

In this paper, a new index-based hybrid checkpointing scheme has been proposed that uses both incremental and communication-induced checkpointing ideas. This scheme offers reduction of the cost of checkpointing and helps in tackling the problem arising due to 'coasting forward'. Failures have been classified as hard and soft. The proposed recovery approaches have been designed for these two types of failures. It has been shown that for soft failures, the recovery approach offers maximum roll-forward.