Skip to Main Content
A high-speed network needs to assign a fixed bandwidth for each connection some where between its mean and peak rates. Most of the time this assigned bandwidth will not handle all the traffic received and creates traffic loss. This paper introduces a new algorithm to avoid network congestion. The algorithm mainly considers online measurements of the relative contents of each buffer in the network. An adaptive bandwidth reallocation is simply done by recalling an evolved neural network. A particle swarm optimizer (PSO) is used to adjust both weights matrix and the number of nodes for the hidden layer providing that input and output layers are fixed at one node (ratio of relative contents and bandwidth proportion respectively). The results are compared with static bandwidth allocation in terms of number of traffic drop.