Congestion control of Poisson distributed ABR applications in ATM networks using neural network

The standard flow control of available bit rate (ABR) services, as proposed by the ATM Forum, is the rate-based congestion control scheme. This congestion control scheme uses the resource management (RM) cell to carry feedback information. In this paper, a neural approach ABR congestion control scheme that uses an artificial neural network to monitor the queue length in each switch, and then a suitable ER value is calculated based on the queue status. This algorithm is called the neural indicates explicit rate (NIER) scheme. This paper also presents an extensive comparison study of the proposed switching scheme and three other schemes. A standard heavily loaded "parking-lot" network is used to test the effectiveness and efficiency of these schemes. Meanwhile, the on-off ABR applications or Poisson distributed ABR services are used. The simulation results show that the proposed algorithm gives good performance.