Hadoop加速比性能模型研究

MapReduce编程模型被看作是Hadoop的核心组件,一直以来都受到广泛的应用和研究,但针对Hadoop云平台的性能研究主要集中在配置调优和分布式算法等方面,缺乏对Hadoop加速比性能模型的理论研究.针对此问题提出了Hadoop云平台加速比性能模型,并分别在同构机群和异构机群架构上对该模型进行了分析.针对同构机群提出了3种I/O负载模型,并指出了其对加速比的影响;指出了异构机群不适合做Hadoop云平台的原因及异构性对Hadoop云平台造成的影响.这些研究对Hadoop云计算平台的搭建及如何提高Hadoop云平台加速比性能有很好的指导作用,提供了理论依据.

Researchingon Hadoop Speedup Performance Model

As a core technology of cloud computing,MapReduce is used and research broadly,however the research on Hadoop is focused on configuration tuning and distributed algorithms,which results in being lacked of research on Hadoop speedup performance model.This paper proposes the Hadoop cloud platforms speedup performance model and makes an analysis of this model in isomorphic clusters and heterogeneous cluster architecture.According to isomorphism cluster,this paper puts forward three kinds of I/O load model,and points out its influence on speedup.Meanwhile,it indicates the reason and impacts of heterogeneous cluster is not suitable to for the Hadoop cloud platforms.The research has a good guide on building Hadoop cloud computing platform and how to improve the Hadoop cloud platform speedup,provides a theoretical basis.