Research On IP Congestion Control Based On Active Queue Management

With the development of military communication network, the next generation communication system need not only to support low/high speed data, image and other non-voice traffic, but also to support multimedia traffic. Under the condition of limited network bandwidth currently, proper network congestion control mechanism must be adopted in order to satisfy the QoS demand of different network traffic. Due to the lack of predictable congestion, the traditional drop tail algorithm could not prevent the congestion collapse. On the other hand, the Active Queue Management (AQM) algorithm could selectively drop or mark the packets before congestion occurring, and make the source decrease the transmission rate, therefore avoid the occurrence of congestion.This paper first analyzes four representative AQM algorithms of RED, RIO, SRED and BLUE along with the traditional drop tail both in theory and simulation, and fully studies the problems and the corresponding reasons existing in the algorithm. Then proposes the nonlinear smoothing algorithm, which can effectively control the average queue length by smoothing the packet loss function of the RED algorithm nonlinear, and have good congestion control ability. Simulation results based on the platform NS show that this algorithm has been significantly improved compared to the RED algorithm in the performance of packet loss ratio, end-to-end delay, throughput and delay jitter. Finally this paper proposes FCURF(Feedback Control Un-Responsive Flow) algorithm based on feedback control theory. The basic idea of the algorithm is that when non-response flow occupies bandwidth maliciously the algorithm can feedback information to the source using ICMP so as to bring down the sending rate of source. Therefore the algorithm can effectively control the bandwidth occupation of non-response flow.