MBFS: a parallel metadata search method based on Bloomfilters using MapReduce for large-scale file systems

摘要: The metadata search is an important way to access and manage file systems. Many solutions have been proposed tackle performance issue of search. However, the existing build a separate index at internal or external system through related data structure database use semantics event-notification method construct structure, utilize sampling-based conduct direct on namespace, face problems high I/O overhead for maintaining consistency between indexes metadata, enormous space storing low accuracy results so on. To address these problems, this paper presents MBFS, fast, accurate lightweight based multi-dimensional Bloomfilters. We create Bloomfilter basis directory entry that can prune sub-trees narrow scope namespace. MBFS capable producing fast answers class complex over after consuming small number disk accesses. residing in does not need additional maintain consistency. consists Bloomfilters which are composed bits, it consumes marginal overhead. Moreover, employs MapReduce speeding up under environment multiple servers. Extensive experiments conducted prove effectiveness MBFS. experimental show achieve excellent only latency, but also with time