Research On Active Queue Management In Routers

The rapid development widely use of the Internet makes it more closely relate to people's daily lives. In order to meet the needs of users, varieties of new network applications and services appeared and developed. With the rapid growth of network traffic, network congestion's repeating affects its performance seriously. Therefore, the implementation of congestion control becomes a very important method to ensure its stable operating as the same as improve its quality of service (QoS).The main protocol used in the Internet today is TCP/IP. And the network traffic flow also is dominated by TCP. The method of researching congestion control is focused and emphasized on the end to end use of TCP congestion control and the use of IP router congestion control. In order to respond to the congestion positively, active queue management (AQM) technology uses a way of packet loss. It can reduce inequities on the allocation of resources and improve not only the network's QoS, but stability. It is the most important way to reduce and avoid congestion. As a good active queue management algorithm, it should not only reduce the packet loss rate and improve throughput, but also make a good prediction of congestion conditions to ensure the performance of network.By researching and analyzing the RED algorithm and related algorithms on AQM, this thesis uses the approach of a similar "S"-type growth curve equation to balance the relationship of promotion and inhibition between the packet loss rate and the average queue length, called the improved RED algorithm using the model of logistic. The experiments show that the algorithm decreases the packet loss rate by about 28% comparing with the RED algorithm, and about 9% with nonlinear RED algorithm. It also has a significant improvement on throughput, especially the network's stability.