Buffer management using genetic algorithms and neural networks

In ATM networks, many control mechanisms were proposed to manage the buffers by introducing control parameters which are adjustable by the network providers. However, it is difficult to adaptively select these control parameters in ATM networks for the traffic environment is much more complicated. We propose a control scheme using genetic algorithms and a neural estimator in the buffer management of an ATM switch. Simulation results demonstrate that even if the traffic environment and the service requirements are dynamically changing, the proposed control scheme is still effective in adaptively selecting control parameters.