The Research Of Congestion Control On Active Queue Management

With the rapid devolvement of the Internet scale, users and traffics, the network congestion has become more serious and complex due to ever-increasing network application types. Congestion control is an important method that can avoid congestion and control or eliminate congestion effectively, which plays a crucial role in stabilizing network, thus, congestion control is always a hot spot in the field of network research. Active Queue Management (AQM) is a Hotpoint in the field of end-to-end congestion control, which takes some prevent measures to avoid congestion by dropping packets selectively, in order to balance low delay of high throughput .This paper focuses on AQM in Congestion control filed, the main works of this paper include:Firstly, the accurate estimation of traffic stream is the basic guarantee of QoS in AQM mechanism. Considering the CARE algorithm sensitivity to parameter configuration of estimation time ,significance lever value and length of capture list, this paper puts forward an improved CARE traffic flow estimation algorithm, which puts error of two estimation results as judge condition and updates the estimation parameters recursively, then reduces the influence caused by unreasonable parameters configuration. Simulation experiments show that average error of the improved algorithm is 4.80%, while that of CARE algorithm is 71.36%, which, therefore, proves that the improved algorithm can reduce errors and improve the accuracy of estimation.Secondly, Active Queue Management becoming a hot point in end-to-end congestion control, as fuzzy control does not rely on precise mathematical model , this paper develops a fuzzy adaptive virtue queue management algorithm (FAVQ), in order to overcome the shortcomings of AVQ, which can not control the length of control queue. Making the errors between current queue length and expected queue length and the errors of queue length changes to be the fuzzy inputs, FAVQ takes the drop probability which updated by fuzzy control module as the strategy of discarding data packets instead of the strategy in AVQ. FAVQ can update the drop probability timely.Thirdly, Because FAVQ algorithm can't update the rules dynamically and Rough Set can find hidden laws effectively, this paper develops a rough fuzzy adaptive virtue queue management algorithm (RFAVQ). By extracting and reducing the new inference rules timely, RFAVQ can update the inference rules in fuzzy control module dynamically, therefore, improves the performance of queue management.Lastly, realize FAVQ and RFAVQ on Network Simulator - NS2. The test shows that FAVQ can control the queue length in expected value range, with a relatively small jitter. In addition, it can keep a balance between throughput and delay, which proves its robustness and stability. RFAVQ can extract, reduce and update the inference rules dynamically. Compared to FAVQ, all performace references have been enhanced to a certain degree in RFAVQ, which proves the timeliness and effectiveness of updating the inference rules.