Effective Congestion Indication for Performance Improvement of Random Early Detection

A congestion avoidance scheme allows a network to operate in the region of low delay and high throughput. Such scheme prevent a network from entering in to congested state. RED(Random Early Detection), is one such congestion avoidance mechanism used for effectively control of congestion. In RED, router uses only the average queue size, as a congestion indicator and the average queue length is insensitive to input traffic load variation. Due to this effective incipient congestion becomes difficult to detect and there is no matching between current queue size and average queue size as in [4]. The present paper deals with these two problems and proposed a way in which packet dropping is not only based on average queue size but also on the rate of change of input. The work which is carried out is to find out significant changes in input rate and use this climbing rate as indication of impending congestion for sources to react quickly. Here we have analyzed the performance of our proposed algorithm using network simulator ns2.