ACR: Automatic checkpoint/restart for soft and hard error protection

As machines increase in scale, many researchers have pre-dicted that failure rates will correspondingly increase. Soft errors do not inhibit execution, but may silently generate incorrect results. Recent trends have shown that soft er-ror rates are increasing, and hence they must be detected and handled to maintain correctness. We present a holis-tic methodology for automatically detecting and recovering from soft or hard faults with minimal application interven-tion. This is demonstrated by ACR: an automatic check-point/restart framework that performs application replica-tion and automatically adapts the checkpoint period using online information about the current failure rate. ACR per-forms an application- and user-oblivious recovery. We em-pirically test ACR by injecting failures that follow different distributions for five applications and show low overhead when scaled to 131,072 cores. We also analyze the interac-tion between soft and hard errors and propose three recovery schemes that explore the trade-off between performance and reliability requirements.