Redesigning the message logging model for high performance

Over the past decade the number of processors in the high performance facilities went up to hundreds of thousands. As a direct consequence, while the computational power follow the trend, the mean time between failures (MTBF) suffered, and it's now being counted in hours. In order to circumvent this limitation, a number of fault tolerant algorithms as well as execution environments have been developed using the message passing paradigm. Among them, message logging has been proved to achieve a better overall performance when the MTBF is low, mainly due to it's faster failure recovery. However, message logging suffers from a high overhead when no failure occurs. Therefore, in this paper we discuss a refinement of the message logging model intended to improve failure free message logging performance. The proposed approach simultaneously removes useless memory copies and reduces the number of logged events. We present the implementation of a pessimistic message logging protocol ...

• Type: Journal
• Year: 2010
• Where: CONCURRENCY
• Authors: Aurelien Bouteiller, George Bosilca, Jack Dongarra
• Keywords: Better Overall Performance, CONCURRENCY 2010, Past Decade, Performance