Network Congestion Control Based On Improved Nonlinear Red Algorirhm

With the rapid development of computer network, The explosive growth in the network data flow take place, the data flow of network congestion problem has become a more and more valued. When the load is too large, but in the network resource i s very limited, it will cause throughput degradation and the network latency increasing and packet substitution rate increasing, which makes the system performance and robustness of serious decline, and network congestion control algorithms become an effec tive way to solve this problem.In order to improve the network quality of service and ensure the correct transmission of data, It proposed a variety of network congestion control algorithm, in which the most representative of active queue management mechanism is Random Early Detection Algorithm. Compared to the Drop Tail passive queue management algorithm, RED queue management mechanism solve the problem of global TCP synchronization, but before the queue of the active queue management mechanism is full, the mechanism start to drop the packet and limit the speed of the transmission side, so it could effectively avoid the queue overflow caused by throughput decreasing, packet loss rate surging, the delay increasing, and low link utilization. But the lacks of the RED algorithm are stability and parameter sensitivity and so on. Therefore, this article detailed study and research about the RED algorithm applied NS₂ to simulate and improve the RED algorithm on ERED, 2SSRED and 3SSRED in Linux system.First, the new algorithm proposes a new nonlinear smoother and the loss probability function of the RED has been adjusted. New algorithm calculated the relationship between probability of loss rate and average queue length by scientific mathematical method, which shows that the relationship probability of loss rate and average queue length is not a simple linear relationship yet a simple curve relationship, but the relationship of a function of similar S-type.Secondly, the entire length of the queue area divided into different regions. It can be simply divided into two areas: economic zone and the negative zone; It can also be more complex divided into three areas: active area, stable area and negative area. In different regions adjustment of the probability function of the loss rate will be different, distribution and processing of the data will be more accurate and rapid.Finally, applying NS₂ to simulate and verification the new algorithm in Linux system. Simulation results show that the three new algorithms in network throughput, packet loss rate compared showed a clear advantage in RED algorithm.