迭代式MapReduce研究进展  

Advances in iterative Map Reduce

作　　者：李金忠[1] 汤鹏杰[1] 夏洁武[1] 谭云兰[1] 

机构地区：[1]井冈山大学计算机科学与技术系,江西吉安343009

出　　处：《计算机工程与应用》2015年第12期123-132,共10页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61163062);江西省教育厅科技计划项目(No.GJJ14561);江西省科技支撑计划项目(No.20122BBG70161);江西省自然科学基金项目(No.2012BAB201038)

摘　　要：迭代计算普遍存在于大数据处理中,而传统的MapReduce不能显式地支持迭代计算。近几年,研究者扩展和改进原始MapReduce,已开发了若干迭代式MapReduce以更好地为大数据处理而支持迭代计算。对迭代式MapReduce编程框架进行综合评述,较详细地阐述了这些研究成果,给出了它们各自的基本思想,并分析了它们各自的特点、优势和不足,且对比了它们所采用的一些技术。对迭代式MapReduce未来的发展趋势进行了展望。Iterative computations are pervasive among big data processing, but the traditional MapReduce cannot explicitly support iterative computation. In recent years, researchers have extended and improved the original MapReduce, and have developed a number of iterative MapReduce to better support iterative computation for big data processing. A comprehen-sive review of iterative MapReduce programming framework is provided. These research achievements are described in detail. Their basic ideas are given. Their characteristics, advantages and disadvantages are analyzed for each framework, and some technologies that have been adopted in these frameworks are compared. Some promising development trends for future research of iterative MapReduce are pointed out.

关 键 词：MAPREDUCE 迭代计算 迭代式MapReduce 并行编程模型 大数据处理 

分 类 号：TP316[自动化与计算机技术—计算机软件与理论]