Impact of VC Merging on Buffer Requirements in ATM Networks

For the implementation of multipoint-to-point connections in ATM, various approaches exist, each with its own advantages and disadvantages. VP-based methods require unique sender identification
but they do not require reassembly in merging points. In contrast, VC-based methods do not require unique sender identification but they do require reassembly in merging points. It is likely that VC merging will be the method of choice as it is scalable and yet relatively simple to implement. One of its drawbacks is the increased output buffer space required at the switches because of packet reassembly at the merging points. This paper investigates the impact of the switch architecture and characteristics on the output buffer space by means of simulation. The results obtained demonstrate that for typical switch architectures, VC merging does not require significant additional buffering compared to VP merging.