Self organizing map based red parameter estimation for congestion avoidance

In data communication networks, congestion avoidance in routers is one of the favorite research topics. In this dissertation, new queue management approaches are proposed on the Random Early Detection (RED) algorithm by monitoring the global congestion status of an autonomous system. In order to view the level of global congestion, a traffic flow is generated between routers and a centralized observation unit. Internet Protocol (IP) routers are specialized in order to send information packets, regarding current output queue lengths, to the observation unit. By this way, it becomes possible to produce a global picture of congestion. A Self Organizing Map (SOM) structure is used by the observation unit, to make predictions on the future congestion behaviour of the network. After some improvements, routers gain the ability to update their RED parameters according to the congestion notifications sent by the observation unit. In this work, the benchmarks of new queue management approaches are investigated by simulations on OPNET Modeler and comparisons with recent queue management technologies: Drop Tail (DT) and RED, are performed. The results for throughput, packet loss, end-to-end delay and delay variation are observed.