End-to-end delay analysis for networks with partial assumptions of statistical independence: invited paper

By accounting for statistical properties of arrivals and service, stochastic formulations of the network calculus yield significantly tighter backlog and delay bounds than those obtained in a purely deterministic framework. This paper proposes a stochastic network calculus formulation which can account for partial assumptions on statistical independence of arrivals and service across multiple network nodes. Scenarios where this can be useful are packet tandem networks with cross traffic and independent arrivals, where identical packet sizes create correlations across the nodes. As an application, the paper investigates the role of partial statistical independence on end-to-end delay bounds in four main scenarios arising by combining assumptions on the statistical independence of arrivals and packet sizes at different network nodes.