High Performance Cache Management for Parallel File Systems

Abstract

AbstractCaching has been intensively used in memory and traditional file systems to improve system performance. However, the use of caching in parallel file systems has been limited to I/O nodes to avoid cache coherence problems. In this paper we present the cache mechanisms implemented in ParFiSys, a parallel file system developed at the UPM. ParFiSys exploits the use of cache, both at processing and I/O nodes, with aggressive pre-fetching and delayed-write techniques. The cache coherence problem is solved by using a dynamic scheme of cache coherence protocols with different sizes and shapes of granularity. Performance results, obtained on an IBM SP2, are presented to demonstrate the advantages offered by the cache management methods used in ParFiSys.KeywordsParallel ApplicationProcessing NodeTotal ThroughputCache SchemeCache CoherenceThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.