摘要
为了减少网络管理系统NBOS中应用统计功能的响应时间,文中采用OpenMP对系统进行了并行改进.首先,指出由于海量数据的统计,NBOS的应用统计功能需要用并行化的方法加以改进.其次,仔细研究流行的两种并行设计模型(共享内存模型和消息传递模型),结合已有的硬件平台,选择OpenMP作为并行设计工具.由于对数据进行了有效地分割和合理地分配,保证了数据处理的高度并行性.最后,通过从不同时间跨度对结果进行分析,发现系统响应时间减少了一半,很大程度提高了系统实用性.
To overcome the slow response of NBOS,OpenMP is used to develop a parallel program to improve it.So,firstly,it proposes that it is the mass data that results in the slow response and parallel methods should be used to improve the performance of NBOS.Secondly,it studies two concurrent design patterns,which include shared memory pattern and message passing pattern,and selects shared memory pattern to develop the program.Because of the effective data partition and the rational allocation,it makes data processing have a high concurrency.Compared with the original under conditions of different time span,the response time is halved.So,the practicability is improved to a large extent.
出处
《微电子学与计算机》
CSCD
北大核心
2008年第9期75-77,共3页
Microelectronics & Computer
基金
国家自然科学基金项目
国家"八六三"计划项目
关键词
并行
OPENMP
共享内存
性能改进
concurrency
OpenMP
memory-shared
performance improvement