Study On RED Improvement Based On ARM Embedded System

With the expansion of users and applications of the Internet, people are paying more and more attention to the Congestion control mechanism. A router-based congestion control mechanism, which is good at both robustness and network throughput, is an important research direction in currently network congestion control aspects. The disadvantages of traditional TCP congestion control mechanisms, such as delay and hysteresis, can be avoided by using Router-based congestion control mechanisms.Random Early Detection(RED),which has been the only candidate of AQM recommended by IETF,monitors the queue length at the gateway,drops/marks the arrived packet randomly before the caches overflow,but in terms of response time,stability and robustness of the network environment is still deficient. Since currently ARM processor that based embedded system platform have used in varuous routers widely,the router that based ARM processor was selected as experimental platform in this paper. According to the few hardware resources and the low computational complexity, an improved RED Algorithm was proposed, which is more suitable for ARM processor that based embedded system.First,this paper introduces the mechanism of the TCP congestion control and the mechanism of the congestion control based on routers,and the design motives,strengths and advantages/disadvantages of RED are detailed analyzed. On basis of it, establishs and model of parameters which suitable for ARM embedded system and proposes an improved program of RED based on throry of random process and self_feedback control. The main research results are as follows:1.This paper improves the mechanism of packet loss probability, which can increase the effectiveness of regulatory mechanisms and improve the utilization and throughput of link because of the Non-linear adjustment.2.The performances of the RED algorithm are greatly affected by the load of network. This paper improves the stability of network through the use of self-adaptive control mechanism which adjusting the maximal drop probability.3.For the RED algorithm is not sufficiently to adapt to the non-punishment stream and not equitable distribution of bandwidth. So, an improved algorithm is proposed in this based on the characteristics of data streams.The improved RED algorithm is verified in Network simulation comparing to the original RED algorithm, and the results show that the improved algorithm excelled to RED algorithm on adaptability, response, fairness, and robustness.In addition,to observe the running performance of the improved algorithm,The contrast experiment was made in which the improved RED algorithm run on both the router that based ARM processor and personal computer platform respectively. Experiment results show that there is no much difference whether the improved RED algorithm run on the router that based ARM processor or personal computer platform. It means the improved RED algorithm is very suitable for running on ARM router that based embedded system.