Performance of access rate control for correlated sources in high-speed networks: Spectral analysis

Abstract

In this paper we study the performance trade-offs of leaky-bucket control in a high speed network with correlated sources. The emphasis is placed on the interrelationship between access control queueing and network queueing, as affected by degrees of control in response to differing input power spectrum characterizing the source correlations. Both stochastic and deterministic analyses are developed. Whereas the stochastic analysis based on multi-state Markov modulated sources provides exact performance measures of the access rate control in random traffic environment, the deterministic analysis based on sinusoidal sources complements the stochastic analysis, leading to good engineering insights into the control performance in the complex scenarios where the stochastic analysis is intractable. Interesting observations are made between the control performance and the low-frequency input power. In the stochastic analysis, we use a novel technique to characterize a rather general class of input power spectrum and are able to derive not only the power spectrum of the leaky-bucket controlled source but also the distribution of a network queue fed by the controlled source.