首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于队列的处理器亲和机制及在网络转发中的应用
引用本文:姚崎,韩臻,高志明.基于队列的处理器亲和机制及在网络转发中的应用[J].北方交通大学学报,2010(2):106-110.
作者姓名:姚崎  韩臻  高志明
作者单位:北京交通大学计算机与信息技术学院,北京100044
基金项目:国家“863”计划项目资助(2007AA01Z410); 电子信息产业发展基金项目资助(2007-329)
摘    要:提出一种新的基于多报文队列的处理器亲和机制,与传统基于网络接口的亲和机制相比,消除了多处理器对共享资源的竞争,优化了处理器的CACHE行为,降低了互斥锁的影响范围,能够有效提升多核处理器系统的网络转发处理能力.试验结果显示,在复杂网络环境下,基于队列的亲和机制报文转发性能较传统基于接口的亲和机制提高了1.8倍.

关 键 词:网络转发  多核处理器  处理器亲和  CACHE行为

Queue-Based Processor Affinity and Its Network Forwarding
YAO Qi,HAN Zhen,GAO Zhiming.Queue-Based Processor Affinity and Its Network Forwarding[J].Journal of Northern Jiaotong University,2010(2):106-110.
Authors:YAO Qi  HAN Zhen  GAO Zhiming
Institution:(School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China)
Abstract:A new affinity scheme based packets queues was proposed.Against the traditional affinity scheme based network interface,the new scheme eliminated the contention of shared resources,optimized the processor CACHE behavior,and reduced the impact of spin-lock.So the packet forwarding capability of multi-processors system was effectively raised.The experiment results shows that the performance of network packets forwarding with the new scheme can be improved 180 percent in the complex network environment.
Keywords:network forwarding  multi-core processor  processor affinity  CACHE behavior
本文献已被 维普 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号