Reconfigurable Nonlinear GRED Algorithm

Abstract

In this paper, a reconfigurable queuing scheme based on Gentle Random Early Dropping (GRED) algorithm has been proposed. Two nonlinear GRED algorithms have been suggested to study the effect of maximum threshold (maxth) and maximum dropping probability (maxp) in the GRED algorithm. In both algorithms, the nonlinearity degree of the two parts of the dropping probability curve of the GRED algorithm is computed to prevent the discontinuity between them. After that, a third algorithm (called Reconfigurable Nonlinear GRED “RNLGRED”) is proposed. This algorithm uses the reconfigurability technique to select the best nonlinearity configuration between the first two proposed algorithms. A simulation carried out in OPNET shows that the proposed RNLGRED algorithm achieves a significant reduction in the queuing delay, jitter, and average queue size when compared with the classical GRED scheme. The reduction in queuing delay, jitter, and average queue size, for example in scenario-1, is up to 40%, 40%, and 35% respectively according to the implemented scenario.