Abstract:
Abstract - In order to make the best utilization of a link in computer networks, Active Queue Management (AQM) algorithms are generally used in a variety of methods. In this paper, a systematic control strategy based on self-scheduled Hp, considering the dynamics of the window-based TCP Row, is proposed. Then, the robustness of the designed controller to disturbance and parameter fluctuations of the plant is shown via simulation results.
Machine summary:
The model is a couplcd nonlinear differential equation as follows: (View the image of this page) Jazu• y/Juue,2It06 Where: W: average TCP window size (packets); q: average queue length; / = + Tp : Round-trip time (secs); Tp: the fixed propagation delay; C: link capacity (packets/sec); N: load factor (number of TCP sessions); p: probability of packet mark/drop.
Considering the time-varying characteristic of parameters it can be shown that the linearized model would be as Figure 1: (View the image of this page) Figure 1 G(s) would be considered as the plant and Kbs) is the controller to assign péi9 in such a way that qLt) well tracks q0 specified by the operator or designer, 8 is the vector [fi C J.
From Figure 5, it is seen that although the parameters are moving dramatically in a wide range according to trajectory of Figure 4, the buffer length has a good track of the reference input, (View the image of this page) ma ni o n Jourii at of' Information Science & Technology, Voi u me 4, Number 1 ,1an ma ry / .
The controller was designed to provide buffer management considering the linearized model of TCP connection dynamics for bottleneck routers, it was very robust to the change of dynamics of the plant caused by random nature of data traffic and was claimed to remove some of the shortcomings .
” ACM SIGCOMM 200a, March 2000.