Analysis of large-scale distributed information systems

Abstract

In this paper we study the effects of correlations between the interarrival times of different service classes. An analysis of distributed information systems reveals that such inter- class correlations exist, in part as a result of the interactions between the server and its clients. To gain insight into the performance implications of these correlations, we formulate a general stochastic model that explicitly captures client-server interactions, and we derive a matrix-analysis analysis of a specific instance of the model. Our results illustrate and quantify the impact that such inter-class correlations can have on system performance.