• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 高性能计算 • 上一篇    下一篇

面向CPU-GPU异构系统的数据分析负载均衡策略

孙婷婷,黄皓,王嘉伦,翁楚良   

  1. (华东师范大学数据科学与工程学院,上海 200062)
  • 收稿日期:2018-09-03 修回日期:2018-11-08 出版日期:2019-03-25 发布日期:2019-03-25
  • 基金资助:

    国家重点研发计划(2018YFB1003400)

A load balancing strategy on heterogeneous
 CPU-GPU data analytic systems

SUN Tingting,HUANG Hao,WANG Jialun,WENG Chuliang   

  1. (School of Data Science and Engineering,East China Normal University,Shanghai 200062,China)
  • Received:2018-09-03 Revised:2018-11-08 Online:2019-03-25 Published:2019-03-25

摘要:

应用于高性能计算领域的通用GPU拥有强大的并行计算能力,以通用GPU作为主处理器的数据分析系统相较于传统数据库能够提供更好的性能。在大数据场景下,如何根据CPU和GPU的资源在处理器之间合理分配工作负载是亟待解决的问题。提出了一种CPUGPU异构数据分析系统上的负载均衡处理策略。该策略采用流水线模型将工作负载分解,基于流水线设计了负载均衡模型,将工作负载合理分配至异构处理器,减少系统总执行时间开销,实现了性能提升。实验结果表明,提出的基于流水线的负载均衡模型能适应不同查询请求下的不同数据量场景,具有良好的性能。

关键词: GPU, 异构负载均衡, 流水线并行, 数据分析处理

Abstract:

With strong parallel computation power, GPU-based data analytic systems can achieve better performance than traditional CPU-based data analytic systems. However, how to leverage the resources of CPU and GPU to dispatch workloads appropriately in massive data scenarios remains to be solved. We propose a load balancing strategy on heterogeneous CPU-GPU data analytic systems. It adopts a pipeline model to split workloads, and a load balancing model based on pipeline is proposed to dispatch workloads to heterogeneous processors reasonably and appropriately, which reduces the total execution time of the system and enhances the performance. Experimental results show that the load balancing model based on pipeline  is suitable for various queries with different dataset volumes and has great performance.
 

Key words: GPU, heterogeneous workload balancing, pipeline parallelism, data analytics and processing