CSFQ算法分析与改进  

Analysis and improvement of CSFQ algorithm

作　　者：李鹏翀[1] 朱荣[2] 饶小兵[1] 秦前清[3] 

机构地区：[1]武汉大学电子信息学院,湖北武汉430079 [2]武汉大学多媒体实验室,湖北武汉430079 [3]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079

出　　处：《计算机工程与设计》2006年第12期2140-2142,2156,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(904160704)

摘　　要：核心无状态公平队列调度(CSFQ)算法提供了如同有状态网那样好的公平带宽分配,但它的丢包算法不适用于TCP流。针对TCP流的特点,对CSFQ算法进行如下改进:将缓存队列长度与丢包概率关联起来,用一种类似于RED(randomearlydrop)缓存管理方法解决了缓存频繁溢出导致的一些问题;对TCP流的丢包率进行修正,使用多余带宽来转发TCP包,解决TCP流与UDP流的带宽分配公平性。仿真试验表明,新算法NEW-CSFQ更好地提供数据流公平的频宽共享,对突发流响应较原算法有所提高,且算法复杂度简单,容易在高速核心路由器上实现。CSFQ （core stateless fair queuing） can achieve as well fairness in bandwidth allocation as that implemented by stateful networks. However, its dropping algorithm is not suit TCP flows. An improved CSFQ algorithm is proposed in consideration of the characters of TCE There are two main improvements in this algorithm. Firstly, the dropping probability is related to the buffer length and a buffer policy similar to RED （random early drop） is used to solve such problems as frequent buffer overflow. Secondly, a new dropping policy which use the extra bandwidth to transfer TCP packets is employed to TCP. Hence, the fairness of bandwidth allocations between TCP and UDP is well achieved. Simulation results show that NEW-CSFQ algorithm provide better flow fair bandwidth share, have more rapid response to busty flows, and its complexity is simple compared with CSFQ. This method is easily utilized in high-speed core routers.

关 键 词：公平队列算法 拥塞控制 传输控制协议 用户数据报协议 无状态公平队列 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]