This paper develops a rather general and easy-to-use methodology how to obtain good approximations for queue length distributions in priority and generalized processor sharing systems where the cumulative input traffic streams are modeled as general Gaussian processes with stationary increments. The idea is to identify the most probable path in the threshold-exceeding event, or a heuristic approximation of it, and then to use probability estimates based on this path. The method is particularly useful for long-range dependent traffic and complicated traffic mixes, which are difficult to handle with traditional queueing theory.
- gaussian processes
- generalized processor sharing
- priority queues
- most probable paths