Research And Implementation Of TFRC Congestion Control Algorithm

Currently, the gaining popularity of multimedia-based applications and itsincreasing use, network congestion starts to be problematic. Due to multipathpropagate, delay, Doppler Effect and fading of wireless network, the followingproblems can be observed: high bit error rate, high packet loss rate, etc. The TCPFriendly Rate Control (TFRC) is a congestion control algorithm which has became agood solution to solve these problems,and also it is a key point of my research.At fist, the thesis discusses the theory of TFRC congestion control algorithm.TFRC shows very good performance with cabled IP networks. Both bandwidthutilization ratio and rate smoothness are good, meanwhile it maintains nice TCPfriendliness. However, traditional TFRC implementations have showed poorperformances for multimedia transmission on wireless networks characterized mainlyin a low utilization rate of available bandwidth. Therefore, the thesis proposes asimple but efficient improved algorithm called TFRC-g. In this improved algorithm,the TFRC connection is virtualized to transfer multi-traditional TFRC flows. At thesame time, the number of parallel TFRC flows is dynamically adjusted according tothe queuing delay of the data network. Through such a mechanism, TFRC-g cansignificantly improve the utilization rate of wireless networks’ bandwidth meanwhilemaintaining good TCP friendliness.The thesis studies TFRC-g congestion control algorithm from theoretical analysisand practice.Firstly, it discusses the difference and relevance between TFRC-g andtraditional TFRC.Then,it illustrates and proofs the use of TFRC-g with a3G mobilevideo alarm system which stably transmits both video and audio on a wirelessnetwork. Finally, the thesis uses NS2network simulation to test the property ofTFRC-g,such as bandwidth utilization ratio, TCP friendliness and so on. Meanwhile,the data results have showed that the TFRC-g congestion control algorithm hasreached the original designed target, been successfully demonstrated with a real software system and will meet the growing demand of wireless networks in mediatransmissions for the next five years.