一种新的实现区分服务的可扩展缓存管理算法  

A Novel Scalable Buffer Management Algorithm for DiffServ

作　　者：魏蛟龙[1] 肖艳华[1] 张驰[1] 

机构地区：[1]华中科技大学电子与信息工程系,武汉430074

出　　处：《电子与信息学报》2004年第5期795-799,共5页Journal of Electronics & Information Technology

基　　金：湖北省自然科学基金(No.2001 ABB104);图像信息处理与智能控制教育部重点实验室开放基金(TKLJ0204)

摘　　要：目前Internet网络中采用的缓存管理策略大多为丢尾(Drop tail)算法,并不能适应区分服务模型的要求;而RED及其改进算法CHOKe也不能提供公平性和相对优先级的保证。该文简要地分析了目前常见的缓存管理算法—RED以及CHOKe所存在的不足,提出了一种基于区分服务模型的新型缓存管理算法—D-CHOKe。该算法提供了公平性和相对优先级的保证,并且保持下现有IP网的可扩展性。仿真试验表明,该算法能够较好地满足区分服务的要求,为不同优先级的流分配不同的带宽。Current Internet uses typically drop tail as its buffer management scheme. This scheme does not meet the need of DiffServ model, while RED and its improved algorithms also suffer from shortcoming in term of fairness and relative drop precedence. This paper analyzes several popular packet scheduling algorithms, proposes a novel scalable algorithm named D-CHOKe(Difference-CHOKe) based on DiffServ model. This algorithm achieves fairness and relative drop precedence, and keep the scalability of the IP network. Simulation demonstrates that D-CHOKe algorithm well meets the requirements of DiffServ.

关 键 词：区分服务 缓存管理 可扩展性 丢弃优先级 D-CHOKe 

分 类 号：TN915.07[电子电信—通信与信息系统]