基于Hash-LRU的超时流管理算法  

Time-Out Flow Management Based on Hash-LRU

作　　者：李海莉[1] 朱珂[1] 李向涛[1] 张校辉[1] 张家齐[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,河南郑州450002

出　　处：《信息工程大学学报》2015年第3期358-363,共6页Journal of Information Engineering University

基　　金：国家863计划资助项目(2011AA01A103)

摘　　要：随着UDP流的增多,传统的流结束标识不再适用于判别流结束,目前常采用流超时策略结束流。分析现有的超时流策略,指出其局限性,提出一种基于Hash-LRU的超时流管理算法。利用Hash算法快速定位分组所属表项并更新,通过有区分的最近经久未用算法(least recent used,LRU)区分活动流和不活动流,活动表项在链表头部,不活动表项在尾端被删除。分析和仿真结果表明,与LRU算法相比,Hash-LRU算法将大流的漏判率降低两个数量级,同时该算法满足流量测量的实时需求。With the continuous increasing of UDP flows, the traditional flow ending identification method no longer applies to the determination of the end of the flow. It is common to end a flow by the time-out flow strategy. This article analyzes the state-of-the-art time-out flow strategies, points out their limitations, and presents a timeout flow management algorithm based on Hash-LRU. The algorithm locates and updates the entry the packet belongs to quickly by using the Hash algorithm, and separates the active flows from the inactive flows by using the distinctiveness LRU algorithm. The active flow entries are in the list head and the inactive ones are in the end and are deleted. A- nalysis and simulated results show that the Hash-LRU algorithm can reduce the missing rate of the elephant flows by two orders of magnitude when compared with the LRU algorithm. The algorithm can meet the needs of real-time flow measurement at the same time.

关 键 词：流量测量 流管理 超时流 LRU 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]