| Self-similarity causes performance degradation in the queueing delay and buffer overflow. To analyze the performance, queueing approaches are adopted to model various types of network nodes including routers, switches, and packet multiplexers. To circumvent the difficulty of exact analysis due to generality in the models' stochastic structures, approximations and bounds have been developed in the past. They, however, yielded the performance estimates that are often grossly inaccurate under the presence of self-similarity.; In this study, we investigated the impact of certain key parameters of self-similar traffic (e.g., Hurst parameter and fractal onset time) on the average queueing delay utilizing experimental queueing analysis. Based on the understanding obtained, we develop approximation equations for the queueing delay under the self-similar arrivals with exponential and constant packet-size distributions. Using the approximations, we develop a prediction-based routing scheme that can be embedded in the link-state routing protocol in real time. The results of our study provide a practical means of finding a lower delay path in data networks under the self-similarity. |
