Study On Active Queue Management Based Congestion Control

The network congestion cannot be avoided since the network's nature is dynamic and decentralized. Neither increasing network resources nor decreasing applications is sufficient to guarantee the stability of the network behaviors and, after all, they are not active ways. So an effective congestion avoidance mechanism is the only way to prevent the congestion collapse from a relative light congestion. The mechanisms of congestion control can be classified into many groups based on different rules. The algorithm of active queue management (AQM) that is applied in the inter node of the network is an active close-loop link algorithm to control congestion. This algorithm generates indicating congestion information in the link node of the network and transmits it to the end system, which is used by the end system to adjust the input rate to alleviate the network congestion. In this thesis, the advantages and disadvantages of the current prevalent congestion control algorithms are discussed in detail. Based on some general AQM algorithms such as RED, REM, RED/ECN and RED/BECN, an improved congestion control scheme is proposed. The main contributions in this thesis are as follows: 1. Proposes a new algorithm—TRED: A TTL Based Active Queue Management algorithm. The algorithm establishes on the famous AQM algorithm-RED and combines ECN and BECN to serve as a congestion notification. By exploiting the advantages of the ECN and BECN and diminishing the disadvantages of them, the algorithm can notify the congestion condition more effectively than previous algorithms. 2. The mathematical model of TRED is constructed in this thesis. Based on this model, the feasibility and the effectiveness of this algorithm are carefully discussed. The analysis results show that it does better than previous algorithm in controlling the congestion and improving the throughput in a congested network. 3. The simulation of the algorithm in the NS (Network Simulator) presents a strengthen support to the results of the theoretical analysis.