【摘要】：MapReduce編程模型被看作是Hadoop的核心組件,一直以來(lái)都受到廣泛的應(yīng)用和研究,但針對(duì)Hadoop云平臺(tái)的性能研究主要集中在配置調(diào)優(yōu)和分布式算法等方面,缺乏對(duì)Hadoop加速比性能模型的理論研究.針對(duì)此問(wèn)題提出了Hadoop云平臺(tái)加速比性能模型,并分別在同構(gòu)機(jī)群和異構(gòu)機(jī)群架構(gòu)上對(duì)該模型進(jìn)行了分析.針對(duì)同構(gòu)機(jī)群提出了3種I/O負(fù)載模型,并指出了其對(duì)加速比的影響;指出了異構(gòu)機(jī)群不適合做Hadoop云平臺(tái)的原因及異構(gòu)性對(duì)Hadoop云平臺(tái)造成的影響.這些研究對(duì)Hadoop云計(jì)算平臺(tái)的搭建及如何提高Hadoop云平臺(tái)加速比性能有很好的指導(dǎo)作用,提供了理論依據(jù).
[Abstract]:MapReduce programming model is regarded as the core component of Hadoop and has been widely used and studied. However, the performance of Hadoop cloud platform is mainly focused on configuration tuning and distributed algorithms. There is a lack of theoretical research on Hadoop speedup performance model. To solve this problem, a speedup performance model of Hadoop cloud platform is proposed, and the model is analyzed on isomorphic cluster and heterogeneous cluster architecture, respectively. Three kinds of I / O load models for isomorphic clusters are proposed, and their effects on acceleration ratio are pointed out, and the reasons why heterogeneous clusters are not suitable for Hadoop cloud platforms and the effects of heterogeneity on Hadoop cloud platforms are pointed out. These studies have a good guidance on how to build Hadoop cloud computing platform and how to improve the speedup performance of Hadoop cloud platform.
【作者單位】： 河南師范大學(xué)網(wǎng)絡(luò)中心;河南師范大學(xué)計(jì)算機(jī)與信息工程學(xué)院;
【基金】：河南省基礎(chǔ)與前沿技術(shù)研究項(xiàng)目(132300410389)
【分類號(hào)】：TP302.7

【參考文獻(xiàn)】

相關(guān)期刊論文 前6條

1 孫健;賈曉菁;;Google云計(jì)算平臺(tái)的技術(shù)架構(gòu)及對(duì)其成本的影響研究[J];電信科學(xué);2010年01期

2 欒亞建;黃爛，

本文編號(hào)：2125235