基于缓存竞争优化的Linux进程调度策略  

Linux Process Scheduling Strategy Based on Cache Contention Optimization

作　　者：夏厦[1] 李俊[1] 

机构地区：[1]中国科学技术大学自动化系,合肥230027

出　　处：《计算机工程》2013年第4期58-61,共4页Computer Engineering

基　　金：中央高校基本科研业务费专项基金资助项目(Wk2100100006)

摘　　要：分析Linux经典内核版本2.6.22的进程调度算法,利用性能监测单元的监测信息,给出3个性能指标CMR、CRR、OCIP对进程的缓存竞争性强弱进行刻画,以此为依据,采用轮询算法优化Linux下的进程调度顺序,尽量避免在CPU上同时运行多个缓存竞争力强的进程,减小系统因缓存竞争产生的性能损耗。在benchmark上的测试结果表明,该方法能够提升系统在中、高负载下运行时的性能,在高负载下运行时的性能提升比例可达6%左右。According to the analysis of the process scheduling algorithm of Linux kemel 2.6.22, through the information getting from Performance Monitor Unit（PMU）, this paper puts forward three performance indexes CMR, CRR and OCIP to describe the process behavior of cache contention. On the basis of this, it uses polling algorithm to optimize the process scheduling in order to reduce the cache contention of Last Level Cache（LLC）. Benchmark test results show that this algorithm can improve about 6% performance when the system load is high.

关 键 词：LINUX内核 片上多处理器 进程调度 性能监测单元 进程行为 缓存竞争 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]