Programmer-directed partial redundancy for resilient HPC

Abstract

In this work we propose partial task replication and check-pointing for task-parallel HPC applications to mitigate silent data corruption (SDC) errors. As the complete replication of all application tasks can be prohibitive due to resource costs, we introduce programmer-directed selective replication mechanism to provide fault-tolerance while decreasing costs. Results show that our scheme detects and corrects around 65% of SDC errors with only 4% overhead on average.Peer ReviewedPostprint (published version

Combining Checkpointing and Replication for Reliable Execution of Linear Workflows

“…Partial redundancy is studied in [13,31,32] (in combination with coordinated checkpointing) to decrease the overhead of full replication. Adaptive redundancy is introduced in [19], where a subset of processes is dynamically selected for replication.…”

Coping with silent and fail-stop errors at scale by combining replication and checkpointing

“…Although partial replication has been empirically studied by some previous work [23,44,45], designing an optimal strategy that combines partial redundancy and checkpointing and analyzing its efficacy remain to be done.…”

“…), and a set of experimental and simulation results. Partial redundancy is studied in [23,44,45] (in combination with coordinated checkpointing) to decrease the overhead of full replication. Adaptive redundancy is introduced in [29], where a subset of processes is dynamically selected for replication.…”

Designing and Modelling Selective Replication for Fault-Tolerant HPC Applications

“…In this work we model HPC application reliability formally by Markov chains and we propose a dynamic runtime heuristic for utilizing idle resources to maximize the reliability of HPC applications. Our work [16] proposes a programmer-guided partial redundancy mechanism for SDCs and fail-stop errors.…”