摘要： 为优化大量云计算线程对共享存储的访问，提出一种共享存储MapReduce云计算性能测试方法。以Oprofile为中心，对云计算的应用逻辑、动态共享库及内核系统调用进行性能统计分析，通过Valgrind与Ptrace机制完成对存储访问及系统调用的计数与计时。实验结果表明，该方法可快速定量分析共享存储的MapReduce，发现应用的内在性能瓶颈。

关键词: 云计算, 共享存储, 性能测试, 系统调用, 动态共享库

Abstract: In order to tune performance of thread-based cloud computing in shared memory, this paper proposes a performance test method for shared memory MapReduce cloud computing. Based on Oprofile, it collects the performance data from application logic dynamic shared library and the kernel system call. Valgrind and Ptrace are used to evaluate memory system and monitor the system calls invoked by application. Experimental results show that the proposed method can analyze the application performance form different levels, and find performance bottleneck in the application.

Key words: cloud computing, shared memory, performance test, system call, dynamic shared library

中图分类号: 

• TP301.6