流级别的高速网络流量动态划分算法  

Dynamic Partitioning of High-speed Network Traffic with Flow Level

作　　者：熊兵[1,2] 肖红光[1] 龙敏[1] 鲁宏伟[2] 

机构地区：[1]长沙理工大学计算机与通信工程学院,长沙410004 [2]华中科技大学计算机科学与技术学院,武汉430074

出　　处：《小型微型计算机系统》2013年第5期945-950,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61001004;61163053)资助;湖南省教育厅科研项目(11C0036)资助;长沙理工大学人才引进基金项目资助

摘　　要：针对高速网络流量并行处理的需求,提出一种流级别的高效动态流量划分算法DTP-PM.讨论了DTP-PM算法设计的基本思想,并分析了其中的关键问题—会话表设计.为提升算法性能,通过定义会话标识符设计了会话表的哈希函数PRH,同时应用MTF启发法改进会话表的哈希冲突解决方法.在引入负载度量化网络流量分析负载轻重程度的基础上,给出流量划分算法DTP-PM的实现流程.借助高速网络流量样本对该算法性能进行了实验评估.实验结果表明,DTP-PM算法在均衡性和高效性方面明显优于经典的流量划分算法.Targeting the requirements of parallel processing of high-speed network traffic, this paper proposes an efficient dynamic traffic partitioning algorithm with flow level, DTP-PM （Dynamic Traffic Partitioning based on PRH-MTF session table）. We firstly discuss the fundamental of the algorithm and conclude one of its essential problems, session table design. To advance the algorithm performance, we devise a pseudo-random hashing （PRH） based on defining session identifier, as the hash function of the session table. Its hash collision resolution is improved by applying move-to-front （MTF） heuristic. Secondly, we describe the implementation of traffic partitioning algorithm DTP-PM based on introducing load degree to quantize the load extent of network trifle analysis. Finally, we evaluate our algorithm by experiments with physical high-speed network traffic. Experimental results indicate that our proposed algorithm DTP-PM outperforms classical traffic partitioning algorithms in terms of balance and efficiency.

关 键 词：高速网络 动态流量划分 流级别 伪随机哈希 MTF启发法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]