Random Early Detection (RED) algorithm was first proposed by Sally Floyd and Van Jacobson in
1) for Active Queue Management (AQM) and then normalised as an IETF recommendation in
2). It is stated that RED is able to avoid global synchronization of TCP streams, maintain a high throughput as well as a low delay and achieve fairness over multiple TCP connections etc.
The introduction of RED has aroused considerable research interest in understanding Its fundamental mechanisms, its performance and the configuration of its parameters to adapt to different working environments. This report first describes the RED algorithm in Section I and then explains several analytical models in Sections II and IV, respectively. Specifically, section II discusses the analytical evaluation of RED functioning, which is based on paper 3). Section IV examines a feedback control model for RED, which was first presented in document 4). In section V, the parameter setting for RED is discussed. The report ends with an additional discussion of selected topics and possible future work. Note that this report only focuses on the original RED algorithm, although numerous RED variants have been proposed.
The RED gateways keep the average queue size low, occasionally allowing packet bursts in the queue. During congestion, the probability that the gateway notifies a particular connection to reduce its window is approximately proportional to the portion of that connection of the bandwidth through the gateway. The RED gateways are designed to accompany a transport layer congestion control protocol, such as TCP. The RED gateway has no bias against bursts traffic and avoids the global synchronization of many connections by decreasing its window at the same time. Simulations of a TCP / IP network are used to illustrate the performance of the RED gateways.