A development of an intelligent algorithm for bandwidth allocation in ATM networks using Petri Nets

In Video on Demand (VoD) service, an algorithm for dynamic bandwidth allocations is essential because the variable bit rate (VBR) of video data streams is inherently bursty. Typical dynamic allocations to competing streams are done in linear proportion to the bandwidth requirements. However, since this scheme has not taken into account statistical nature of compressed video data and constraints of buffers, it is not suitable to directly implement in the scheduler. We propose an algorithm that incorporates fair bandwidth allocation, statistical property of compressed video data, and constraints of buffers. First, we describe the constraints imposed by the buffer to avoid underflow or overflow. Based on these constraints, a new scheme is then designed to ensure a much fairer bandwidth allocation. Finally, the proposed scheme is verified with video trace via ExSpect 6.1 based on Petri Nets.