Modeling And Simulation Of A New Active Queue Management Algorithm Autored Achieve

The last 10 years have witnessed a rapid development of Internet. Since existing Internet connections are failing to fully satisfy the users' increasing demands, network congestions are unavoidable. Recent study reveals that the queue management mechanism of routers plays an important role in shaping the performance of TCP connections. In 1998, Active Queue Management (AQM) was introduced by B. Braden at el as an end-to-end congestion control method to reduce queue delays and maintain system throughput. Later, various AQM techniques were proposed, such as RED. However, most of them turn out to suffer from some limitations when applied to real networks. The primary goal of this paper is to explore a new and feasible Active Queue Management algorithm to improve the network capability. During our research, we need to analyze its performance by observing results from simulations and experiments. The author has achieved the goal above. In this paper, a network model with Random Early Detection technique is proposed and studied. After theoretical analysis and simulating using MATLAB, the instability of the network is pointed out, which is brought by the mutual influence of TCP/IP flow control mechanism and Random Early Detection when a close loop with feedback is formed. A new AQM technique called AutoRED is then proposed to solve this problem. The feasibility and stability of AutoRED is studied theoretically. Parameters of AutoRED are also discussed in this paper. What's more, AutoRED is implemented in NS-2. Two experiments are designed and carried out using RED and AutoRED respectively to compare their performance. With the simulation results, it is proved that AutoRED is able to reduce jitters in the network and has good stability and robustness even under heavy network loads. Although there exists many AQM techniques, none of them could applied to all networks and dozen of congestion related problems has not been settled either. So in face of increasing network congestions, AutoRED could be a useful complementarity of existing congestion control mechanism and has academic and realistic value.