An efficient recovery technique for distributed systems

摘要

In this paper, an index-based checkpointing/recovery approach is proposed. It uses both basic and communication-induced checkpoints. The 'direct-dependency concept' is implemented to design an efficient recovery algorithm by piggybacking only one integer with each application message. The main feature of the algorithm is that all the participating processes run the algorithm simultaneously. Therefore it reduces the time overhead to a good extent. Besides, the algorithm avoids some unnecessary comparisons while determining a consistent global checkpoint; thereby, guaranteeing less computational overhead.