Realtime ATM Traffic Generation

The evolution of new networking protocols has shown, that the conventional traffic models like Poisson are in many cases not suited for simulating real network behaviour [4]. This is especially true for ATM networks, because they will show a high complexity in their data rate dynamic due to statistical multiplexing of data, audio and video traffic. On the other hand realistic traffic models are necessary for many network engineering tasks like buffer dimensioning, quality of service estimation and resource planning. In this paper we propose an advanced approach to ATM traffic generation. We make extensive use of parallel processing in order to apply more realistic traffic models in realtime. The main components of the implementation concept are a standard Sun Sparc Station and a custom built ATM hardware. Our concept enables the user to change or modify the traffic model without having to deal with details of the ATM protocol processing or cell generation. Two examples of realtime traffic generators for source and link level traffic are demonstrated: an MPEG cell stream generator and a generator for self-similar ATM traffic.