Performance of non-homogeneous communication networks with Poisson arrivals and dynamic bandwidth allocation

Abstract

Communication network models play a dominant role in analysing several communication systems. It is customary to consider that the arrival of packets to the network follows a Poisson process, but in many communication systems the arrival of packets depends on time and thus makes the process non-homogenous. To incorporate this feature, the communication network models are developed and analysed as a two node tandem communication network with time dependent arrival rates. It is further assumed that the transmission of packets is dependent on the content of the buffer. After allocating the bandwidth through statistical multiplexing the performance of the communication network is evaluated by deriving the explicit expressions for the performance measures like the average number of packets in the buffers, the mean waiting time of packets in the buffer, the throughput of the transmitters, etc. It is observed that the performance of the proposed communication model is better than the earlier models for the specific values of the packetisation. This network is very much useful in ATM scheduling techniques and computer communications.