一种改进的随机早期检测算法

Enhanced Random Early Detection Algorithm

针对随机早期检测算法(RED)参数配置困难的问题,提出了一种自适应阈值RED(SATRED)主动队列管理算法.该算法根据网络拥塞状况,周期性地自动调整最大阈值和最小阈值,以便改善RED路由器的性能.在调整周期内,每到达一个数据包,将平均队列长度和阈值进行比较,在周期结束时根据比较结果对阈值进行调整.仿真结果表明,在路由器缓冲区有限的情况下,相对于简单的尾部丢弃(tail drop)和RED,该算法能为TCP连接提供更好的公平性.

Random early detection(RED) is an effective algorithm for active queue management.To overcome the difficulty in right parameters determination, resulted from quickly variable network conditions of the conventional RED, an enhanced RED algorithm called selfadaptive thresholds RED ( SATRED) is proposed, which can adjust the thresholds according to the congestion condition of network periodically. During each period, average length of routing queue is compared with thresholds on every packet's arriving, and then thresholds are adjusted at the end of the period. The simulation experiments indicate that the proposed algorithm provides better fairness for TCP connections than RED and taildrop.