Selective Early Detection (SED) Queue Management Algorithm: A Modified Version of (RED)

Abstract

A new scheduling module, the Selective Early Detection (SED) mechanism is proposed in this paper. An experimental evaluation in order to demonstrate how this algorithm succeeds in enhancing the quality, improving the performance and easing the deployment of a production level service is carried out. SED algorithm is useful in controlling the Actual average queue size, and hence the queuing delay, with or without a cooperating transport protocol. This control of queuing delay is extremely helpful for delay sensitive real-time traffic. This paper proposes an algorithm based on SED mechanism, a modification of the Random Early Detection (RED) algorithm, and presents a method of improving the Quality of Service (QoS) of RED over IP network infrastructure and with no changes to the existing network. Firstly, a review of RED algorithm works, implemented in IP networks is conducted. Advantages and disadvantages of the proposed algorithm are summarized. Secondly, SED algorithm bounds the queuing behavior by limiting the actual average queue size. Queuing delay that real-time traffic will experience within a router is known that can help network service providers predict whether or not the required QoS can be met. The proposed model is evaluated using Network Simulator (NS-2) version 2.27 over a wide range of test conditions. From NS-2 simulation results, SED improves the throughput and its quality, reduces the end to end delays which includes the processing time of intermediate nodes. This will improve both network and application performance. In conclusion, the use of SED provides a solution to quality degradation of real-time traffic under congested network conditions.