跳到主要导航 跳到搜索 跳到主要内容

Online makespan minimization in MapReduce-like systems with complex reduce tasks

  • Taibo Luo
  • , Yuqing Zhu
  • , Weili Wu
  • , Yinfeng Xu
  • , Ding Zhu Du
  • Sichuan University
  • California State University Los Angeles
  • Taiyuan University of Technology
  • University of Texas at Dallas
  • Ton Duc Thang University

科研成果: 期刊稿件文章同行评审

19 引用 (Scopus)

摘要

In the MapReduce processing, since map tasks output key-value pairs, and reduce tasks take the pairs output by the map tasks and compute the final results. Therefore, reduce tasks are unknown until their map tasks are finished. Also, we assume that map tasks are preemptive and parallelizable, but reduce tasks are non-parallelizable. With these assumptions, we study the scheduling of minimizing makespan. Both preemptive and non-preemptive reduce tasks are considered. We prove that no matter if preemption is allowed or not, any algorithm has a competitive ratio at least 2-1h, we then give two optimal algorithms for these two versions.

源语言英语
页(从-至)271-277
页数7
期刊Optimization Letters
11
2
DOI
出版状态已出版 - 1 2月 2017
已对外发布

学术指纹

探究 'Online makespan minimization in MapReduce-like systems with complex reduce tasks' 的科研主题。它们共同构成独一无二的学术指纹。

引用此