Title :
Enhancement to RED algorithm
Author :
Fengyuan, Ren ; Yong, Ren ; Xiuming, Shan
Author_Institution :
Dept. of Electron. Eng., Tsinghua Univ., China
Abstract :
RED (random early detection) has gained rather more attention as the AQM (active queue management) recommendation algorithm. The instability and difficulty in tuning control parameter settings are two great obstacles. We first identify several design problems that lead to the oscillation occurring in the steady state, such as, the control law is too simple to overcome the effect caused by the queue system inertia, then introduce the varying speed of average queue size as a new state variable to predict the trend of the network load condition, and modify the dropping function using the new variable as enhancement to the standard RED algorithm. The simulation results verify the effectiveness of the enhanced algorithm. The enhancement makes the RED constrain the oscillation and become more robust and adaptable.
Keywords :
Internet; computer network management; protocols; queueing theory; telecommunication congestion control; telecommunication traffic; AQM; IP; Internet protocol; RED algorithm; active queue management; control parameter settings; dropping function; instability; network load condition; queue system inertia; random early detection; recommendation algorithm; state variable; Algorithm design and analysis; Control systems; Engineering management; Protocols; Queueing analysis; Robustness; Size control; Steady-state; Tuning; Web and internet services;
Conference_Titel :
Networks, 2001. Proceedings. Ninth IEEE International Conference on
Print_ISBN :
0-7695-1187-4
DOI :
10.1109/ICON.2001.962310