Research on Linux network packet buffer recycling toward multi-core processor
|更新时间:2024-10-14
|
Research on Linux network packet buffer recycling toward multi-core processor
Vol. 30, Issue 9, Pages: 102-108(2009)
作者机构:
北京交通大学信息安全体系结构研究中心
作者简介:
基金信息:
DOI:
CLC:TP393.08
Published:2009
稿件说明:
移动端阅览
YAO Qi, LIU Ji-qiang, HAN Zhen, et al. Research on Linux network packet buffer recycling toward multi-core processor[J]. 2009, 30(9): 102-108.
DOI:
YAO Qi, LIU Ji-qiang, HAN Zhen, et al. Research on Linux network packet buffer recycling toward multi-core processor[J]. 2009, 30(9): 102-108.DOI:
Research on Linux network packet buffer recycling toward multi-core processor
摘要
提出一种多核架构下Linux网络报文缓冲区重用无锁算法。通过建立Linux网络转发模型
分析了Linux网络报文缓冲区重用机制的工作原理
指出其在多核处理器上扩展性不好的主要原因在于:互斥机制占用了较多的处理时间
较差的Cache行为特征降低了指令执行效率。设计并实现了MSRQ回收重用算法
实现无需互斥机制对重用队列的并行操作
并且具有更好的Cache行为特征。实验证明
MSRQ算法的小包转发性能比原有的缓冲区重用算法提高了67%。
Abstract
A lock-free algorithm for Linux packet buffer recycling on multi-core processor was proposed.Through making a model of Linux network packet forwarding
the principle of Linux packet buffer recycling was analyzed
and the main reason that the mechanism wasn’t extensible on multi-core processor was pointed out .It was that mutual exclusion and cache coherence operations consumed too much CPU clock tick.A new buffer recycling algorithm named MSRQ was designed and implemented.With the new algorithm
the parallel operations on the recycling queue were permitted without mutual exclusion on multi-core processor.And the new algorithm had better cache behavior than original recy-cling algorithm.The experiment result shows that the performance of Linux network packet forwarding with the new algorithm can be improved more than 67%.