The Markov chain model of RED active queuing management algorithm

A novel Markov-chain model of RED is proposed by the authors in the article. The paper gives a detailed description of the model and some preliminary results for exponential and self-similar traffic. The Markov chain analysis results are verified by a simulation model. 1. The RED (Random Early Drop) active queuing management algorithm The majority of nowadays network technologies is based on transmitting the packets. In packet-switched network the data are transmitted using store and forward algorithm – the datagrams are stored in the network nodes and forwarded to the next node when the link is available. The packets are stored in the buffers of limited size, usually in the memory of the network equipment. For each of the buffers there is a queuing management algorithm, which decides, whenever a packet should be kept in the buffer, sent to the outbound link or dropped. The most common queuing algorithm is the FIFO policy. There are passive and active queuing management algorithms. The passive policies drop the incoming packets if, and only if the buffer is full. The active queuing managers (AQM) drop packets with some probability even if the buffer is partially empty. In this way they try to foreseen the network overload, using the fact, that in most of the bandwidth regulation methods in the internet considers the packet lost as a signal of the overload and decrease the datarate. The most well-known AQM is the Random Early Drop algorithm, presented in [2]