Traffic management for the Gentle Random Early Detection using discrete-time queueing

Since some network connections transmit packets larger than others, there could be unbalanced Share in the available bandwidth and buffer rooms among the existing connections. Further, connections that send packets more than others build more buffer spaces, which therefore cause the network to become congested. This paper proposes a discrete-time queueing analytical model based on Gentle Random Early Detection (GRED) method, which controls the congestion incident by decreasing the connections transmitting rates linearly at certain levels. We compare our analytical model with the classic GRED in terms of several different peformance metrics including, average queue length, throughput, overage queueing delay and packet loss rate, in order to identify the one that offers better Quality of Service (QoS). Furthermore, we compare both models according to the packet dropping probability to determine the one that drops smaller number of packets. Lastly in this paper, we investigate the minimum threshold optimal. position at both the GRED and our analytical model router buffers that gives a satisfied performance.