作　　者： ; ; ; ; ;

机构地区： 湖南大学信息科学与工程学院软件学院

出　　处： 《计算机科学》 2004年第5期41-43,162,共4页

摘　　要： 随机早期检测算法(RED)在产生丢包时简单地丢弃到达的数据包,由于Web流占据了网络上75％左右的带宽,而Web流的特点是数据包比较少,这就使得早期丢包的效果难以体现。本文提出一种选择性丢包机制,引进一个类似公平队列的虚队列,检测发送速率过大的流,筛选出丢包的候选链接,拥塞时丢弃候选链接在队列中的第一个连接的数据包。模拟实验结果表明,与队头丢包相比,采用选择性丢包的RED算法可进一步降低网关的丢包率,提高吞吐率,缩短Http的响应时间,提高队列的稳定性。 Random Early Detection (RED) drops the arriving packet when it is necessary to drop a packet, but the Web traffic is the main traffic in the network and is has few packets,thus the effectiveness of the advance drop is not obviously- This paper proposes a selective dropping policy which uses a virtual queue resemble to the Fairness Queue to detect the flows sending more packet more than the fair share and drops the packets from those flows when congested. The simulation results have demonstrated that the RED algorithm with selective drop has decreased the dropping rate .increased the throughput of the gateway,cut down the response time of Http application and improved the stability of the RED queue.

关 键 词： 选择性丢包机制 随机早期捡测算法 数据包 拥塞控制 队列管理

领　　域： [自动化与计算机技术] [自动化与计算机技术]