The HiBench benchmark suite Characterization of the MapReduce-based data analysis

The HiBench benchmark suite Characterization of the MapReduce-based data analysis

ID:40103894

大小:568.68 KB

页数:11页

时间:2019-07-21

The HiBench benchmark suite Characterization of the MapReduce-based data analysis_第1页
The HiBench benchmark suite Characterization of the MapReduce-based data analysis_第2页
The HiBench benchmark suite Characterization of the MapReduce-based data analysis_第3页
The HiBench benchmark suite Characterization of the MapReduce-based data analysis_第4页
The HiBench benchmark suite Characterization of the MapReduce-based data analysis_第5页
资源描述:

《The HiBench benchmark suite Characterization of the MapReduce-based data analysis》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、TheHiBenchBenchmarkSuite:CharacterizationoftheMapReduce-BasedDataAnalysisShengshengHuang,JieHuang,JinquanDai,TaoXie,andBoHuangIntelChinaSoftwareCenter,Shanghai,P.R.China,200241{shengsheng.huang,jie.huang,jason.dai,tao.xie,bo.huang}@intel.comAbstract—TheMapReducemodelis

2、becomingprominentforthecloud.Inaddition,manynewsystemsbuiltontopofHadooplarge-scaledataanalysisinthecloud.Inthispaper,wepresent(e.g.,Pig[3],Hive[4],Mahout[5]andHBase[6])havethebenchmarking,evaluationandcharacterizationofHadoop,emergedandbeenusedbyawiderangeofdataanalys

3、isanopen-sourceimplementationofMapReduce.Wefirstapplications.introduceHiBench,anewbenchmarksuiteforHadoop.ItTherefore,itisessentialtoquantitativelyevaluateandconsistsofasetofHadoopprograms,includingbothsyntheticmicro-benchmarksandreal-worldHadoopapplications.Wethenchar

4、acterizetheHadoopframeworkthroughextensiveevaluateandcharacterizetheHadoopframeworkusingbenchmarking,soastooptimizetheperformanceandtotalHiBench,intermsofspeed(i.e.,jobrunningtime),throughputcostofownershipofHadoopdeployments,andtounderstand(i.e.,thenumberoftaskscomple

5、tedperminute),HDFSthetradeoffsofnewcomputersystemdesignsforthebandwidth,systemresource(e.g.,CPU,memoryandI/O)MapReduce-baseddataanalysisusingHadoop.Unfortunately,utilizations,anddataaccesspatterns.existingHadoopbenchmarkprograms(e.g.,GridMix[7]andI.INTRODUCTIONtheHivep

6、erformancebenchmark[9])cannotproperlyThetransitiontocloudcomputingisadisruptivetrend,evaluatetheHadoopframeworkduetothelimitationsintheirwheremostuserswillperformtheircomputingworkbyrepresentativenessanddiversity.Forinstance,Yahoohasaccessingservicesinthecloudthroughth

7、eclients.Thereareresortedtothesimplisticsortingprograms[10]toevaluatedramaticdifferencesbetweendeliveringsoftwareasaservicetheirHadoopclusters[11].inthecloudformillionstouse,versusdistributingsoftwareasInthispaper,wefirstproposeHiBench,anew,realisticandbitsformillionst

8、orunontheirPCs.Firstandforemost,comprehensivebenchmarksuiteforHadoop,whichconsistsservicesmustbehighlyscalable,storin

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。