Optimal Checkpointing Period with Replicated Execution on Heterogeneous Platforms

Abstract

In this paper, we design and analyze strategies to replicate the execution of an application on two di erent platforms subject to failures, using checkpointing on a shared stable storage. We derive the optimal pa ern size W for a periodic checkpointing strategy where both platforms concurrently try and executeW units of work before checkpointing. e rst platform that completes its pa ern takes a checkpoint, and the other platform interrupts its execution to synchronize from that checkpoint. We compare this strategy to a simpler on-failure checkpointing strategy, where a checkpoint is taken by one platform only whenever the other platform encounters a failure. We use rst or second-order approximations to compute overheads and optimal pa ern sizes, and show through extensive simulations that these models are very accurate. e simulations show the usefulness of a secondary platform to reduce execution time, even when the platforms have relatively di erent speeds: in average, over a wide range of scenarios, the overhead is reduced by 30%. e simulations also demonstrate that the periodic checkpointing strategy is globally more e cient, unless platform speeds are quite close.

Combining replication and checkpointing redundancies for reducing resiliency overhead

“…As each of the basic redundancies has its own drawbacks and merits, some researchers have attempted to obtain the appropriate combination of those redundancies . In , based on combining replication and resubmission redundancies, the Resubmission Impact ( RI ) heuristic is proposed.…”