基于用户公平的随机早期检测算法改进

边界路由器通常为多个接入用户提供访问Internet的共享上行链路，当所有用户的接入带宽大于上行链路带宽时，可能产生拥塞。在这种典型接入方式下，存在基于用户的公平性问题：一旦边界路由器发生拥塞，各接入用户在共享链路上所获得的吞吐率应该与接入带宽成正比。对不同接入带宽用户的吞吐率进行了理论分析和仿真实验，结果表明，随机早期检测算法(RED)不能提供较好的用户公平性服务，针对这一问题，对RED算法的丢包概率的计算进行了改进：在计算某一个包的丢弃概率时，根据其输入接口的链路带宽进行修正。仿真结果表明，改进的算法在解决基于用户公平性方面具有良好效果。

Modifying RED To Provide Fairness For Users

In general, enterprise and other users share the up-link to access Internet in edge router. Congestion would be occurred when summation bandwidth of all access users is larger than that of up-link. This paper proposes the issue of fairness for users which requires throughput for each user should proportional to their access bandwidth. However, our analytical and simulations results show that the random early detection router(RED router) cannot provide the fairness for users. To solve the problem, we modify the calculation of dropping probability in RED according to each packet's input link bandwidth and average bandwidth. Simulation results illustrate that our modification can provide good fairness for users.