IMPACT OF CHECKPOINT LATENCY ON OVERHEAD RATIO OF A CHECKPOINTING SCHEME

Checkpointing reduces loss of computation in the presence of failures. Two metrics characterize a checkpointing scheme: checkpoint overhead and checkpoint latency. This paper shows that a large increase in late ncy is acceptable if it is accompanied by a relatively small reduction in overhead. Also, for equidistant checkpoints, optimal checkpoint in terval is shown to be typically independent of checkpoint latency.