Research Trend of Electrode Materials for Lithium Rechargeable Batteries

Abstract

Big-data systems are increasingly used in many disciplines for importanttasks such as knowledge discovery and decision making by processinglarge volumes of data. Big-data systems rely on hard-disk drive (HDD)based storage to provide the necessary capacity. However, as big-dataapplications grow rapidly more diverse and demanding, HDD storagebecomes insufficient to satisfy their performance requirements. Emergingsolid-state drives (SSDs) promise great IO performance that can beexploited by big-data applications, but they still face seriouslimitations in capacity, cost, and endurance and therefore must bestrategically incorporated into big-data systems. This paper presentsBigCache, an SSD-based distributed caching layer for big-data systems.It is designed to be seamlessly integrated with existing big-datasystems and transparently accelerate IOs for diverse big-dataapplications. The management of the distributed SSD caches in BigCacheis coordinated with the job management of big-data systems in order tosupport cache-locality-driven job scheduling. BigCache is prototyped inHadoop to provide caching upon HDFS for MapReduce applications. It isevaluated using typical MapReduce applications, and the results showthat BigCache reduces the runtime of WordCount by 38% and the runtimeof TeraSort by 52%. The results also show that BigCache is able toachieve significant speedup by caching only partial input for thebenchmarks, owing to its ability to cache partial input and itsreplacement policy that recognizes application access patterns.