Mahdi Jalili-Kharaajoo
Young Researchers Club
Azad University, Tehran, IRAN
P.O. Box: 14395/1355, Tehra, Iran
email: mahdijalili@ece.ut.ac.ir
Abstract:
TCP congestion control mechanism, while necessary and powerful, are not sufficient to provide good service in all circumstances, specially with the rapid growth in size and the strong requirements QoS support, because there is a limit to how much control can be accomplished at end system. It needed to implement some measures in the intermediate nodes to complement the end system congestion avoidance mechanisms. Active Queue Management (AQM), as one class of packet dropping/marking mechanism in the router queue, has been recently proposed to support the end-to-end congestion control in the internet. It has been a very active research area in the internet community. The goals of AQM are (1) reduce the average length of queue in routers and thereby decrease the end-to-end delay experimented by packets, and (2) ensure the network resources to used efficiently by reducing the packet loss that occurs when queues overflow. AQM highlights the tradeoff between delay and throughput. By keeping the average queue size small, AQM will have the ability to provide greater capacity to accommodate nature-occurring burst without dropping packets, the same time, reduce the delays seen by flow, this is very particularly important for real-time interactive applications. RED was originally proposed to achieve fairness among sources with different burst attributes and to control queue length, which just meets the requirements of AQM. However, many subsequent studies verified that RED is unstable and too sensitive to parameter configuration, and tuning of RED has been proved to be a difficult job.
The intuition and heuristic design is not always scientific and reasonable under any conditions. Of course, since Internet is a rather complex huge system, it is very difficult to have a full-scale and systematic comprehension, but importance has been considerably noted. The mathematical modeling of the Internet is the first step to have an in-depth understanding, and the algorithms designed based on the rational model should be more reliable than one original from intuition. In some of the references, the nonlinear dynamic model for TCP flow control has been utilized and some controllers like PI and Adaptive Virtual Queue Algorithm have been designed for that. Some authors have been constructed a unified framework, the problem of congestion control was formulated as a convex program, with the aggregate source utility being maximized subject to bandwidth constrain. Although PI controller successfully related some limitations of RED, for instance, the queue length and dropping/marking probability are decoupled, whenever the queue length can be easily controlled to the desired value; the system has relatively high stability margin. The shortcomings of PI controller are also obvious. Thirdly, the status of actual network is rapidly changeable, so we believe that it is problematic and unrealistic, at least inaccurate, to take the network as a linear and constant system just like the designing of PI controller. Affirmatively, the algorithm based on this assumption should In order to evaluate the performance of the proposed fuzzy controller, some simulations are provided. In these simulations, the set-point tracking problem of queue will be considered. Comparing the performance of the proposed fuzzy controller with that of PI one, we will see that the fuzzy active queue management system is capable of being against the disturbances, and adapting to highly variability and uncertainty in network. The fuzzy controller has the superior steady and transient performance, exhibits the great adaptability to the variances of link delay and capacity, and provides more robustness against the noise and disturbance.