Numerical queueing analysis of loss priority for large buffer in ATM networks

In this paper, a new method to analyze the cell loss behavior of partial buffer sharing priority mechanism is proposed. The traffic model is based on the fluid queue and normal stochastic process and the analytical technique is simple and general, only requiring mean arrival rate and autocovariance function. Some typical sources (such as voice, video and heavy tail sources) are applied to the proposed method. Performance results obtained show that the method is valid and useful. The proposed method has the merits that it can easily solve cell loss probability especially with large buffer and avoid huge state space problem comparing with classical state transition matrix method. In particular, the computational gain is substantial regardless of buffer size. Also, the traffic sources can be homogeneous and heterogeneous, bursty and correlated.