66 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 40, NO. 1, FEBRUARY 2010 Random Early Detection for Congestion Avoidance in Wired Networks: A Discretized Pursuit Learning-Automata-Like Solution Sudip Misra, Member, IEEE, B. John Oommen, Fellow, IEEE, Sreekeerthy Yanamandra, and Mohammad S. Obaidat, Fellow, IEEE Abstract—In this paper, we present a learning-automata-like 1 (LAL) mechanism for congestion avoidance in wired net- works. Our algorithm, named as LAL Random Early Detection (LALRED), is founded on the principles of the operations of ex- isting RED congestion-avoidance mechanisms, augmented with a LAL philosophy. The primary objective of LALRED is to optimize the value of the average size of the queue used for congestion avoidance and to consequently reduce the total loss of packets at the queue. We attempt to achieve this by stationing a LAL algorithm at the gateways and by discretizing the probabilities of the corresponding actions of the congestion-avoidance algorithm. At every time instant, the LAL scheme, in turn, chooses the action that possesses the maximal ratio between the number of times the chosen action is rewarded and the number of times that it has been chosen. In LALRED, we simultaneously increase the likelihood of the scheme converging to the action, which minimizes the number of packet drops at the gateway. Our approach helps to improve the performance of congestion avoidance by adaptively minimiz- ing the queue-loss rate and the average queue size. Simulation results obtained using NS2 establish the improved performance of LALRED over the traditional RED methods which were chosen as the benchmarks for performance comparison purposes. Index Terms—Average queue size, discretized pursuit learning, queue loss, random early detection (RED), stochastic learning automata (LA). I. I NTRODUCTION O NE OF THE main advantages that wired networks offer is their higher degrees of reliability and better connection Manuscript received January 29, 2009; revised August 2, 2009. Current version published October 30, 2009. This paper was presented in part at the Proceedings of AICCSA’09, the 2009 ACS/IEEE International Conference on Computer Systems and Applications, Rabat, Morocco, May 2009. This work was supported in part by the Department of Science and Technology, Govern- ment of India, under Grant SR/FTP/ETA-36/08. This paper was recommended by Associate Editor G. Papadimitriou. S. Misra is with the School of Information Technology, Indian Institute of Technology, Kharagpur 721 302, India (e-mail: sudipm@iitkgp.ac.in). B. J. Oommen is with the School of Computer Science, Carleton University, Ottawa, ON K1S 5B6, Canada, and also with the University of Agder, 4876 Grimstad, Norway (e-mail: oommen@scs.carleton.ca). S. Yanamandra is with the School of Information Technology, Indian Institute of Technology, Kharagpur 721 302, India, and also with Kalinga Institute of Industrial Technology, Bhubaneswar 751 024, India (e-mail: sreekeerthy46@gmail.com). M. S. Obaidat is with the Department of Computer Science, Monmouth Uni- versity, West Long Branch, NJ 07764 USA (e-mail: obaidat@monmouth.edu). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TSMCB.2009.2032363 1 The reason why the mechanism is not a pure LA, but rather why it yet mimics one, will be clarified in the body of this paper. strength as compared to their wireless counterparts. However, the performance of wired networks often degrades to a great extent due to congestion in the network. The latter results in an increase in the packet loss and a corresponding significant decrease in the throughput. Although congestion cannot be curbed permanently (since it is determined by the traffic pat- terns and not by the traffic-routing mechanisms), its adverse effects can be minimized by decreasing the packet drops in the network. Despite the fact that the transmission control protocol (TCP) supports mechanisms such as Slow Start, Congestion Avoidance, and Fast Retransmit and Fast Recovery to decrease the effect of packet loss due to congestion, they are not very effective in curbing down congestion per se [4]. Consequently, we believe that alternative congestion-avoidance mechanisms are needed. To avoid congestion in networks, researchers have ad- vocated the use of active-queue-management (AQM) strate- gies, in which packets are dropped before the queue gets full. Many AQM techniques, such as the adaptive virtual queue, random early detection (RED), random exponential marking, PI controller, and the blue and stochastic blue [11] schemes, have been reported. Among these existing schemes, RED is one of the most widely used techniques in practice. Philosophically, RED is a congestion-avoidance algorithm. This is because it foresees (or anticipates) the congestion by monitoring the average queue size. It also avoids global syn- chronization by randomly choosing packets to be marked or dropped before the queue gets full. The performance of RED is known to be sensitive to its parameters such as the MAXimum threshold (MAX th ), the MINimum threshold (MIN th ), the Maximum packet-marking probability (PMP) (Max P ), and the so-called weighting factor [6], [7]. Before we proceed, we clarify how these parameters affect RED. Let Avg de- note the average queue size. Then, we have the following conditions. 1) If Avg < MIN th , then no packet drops and marks occur. 2) If Avg > MAX th , then all the packets are marked. 3) If MIN th < Avg < MAX th , then the packets are randomly marked with a certain probability whose value varies from zero to Max P , evaluated using (2). 4) Let p b be an intermediate PMP given by p b ← Max P × Avg − MIN th MAX th − MIN th . (1) 1083-4419/$26.00 © 2009 IEEE Authorized licensed use limited to: UNIVERSITY OF AGDER. Downloaded on May 19,2010 at 07:42:07 UTC from IEEE Xplore. Restrictions apply.