📝 SummarXiv

Abstract

Consider a (logical) link between two distributed data centers with available bandwidth designated for both deadline-driven elastic traffic, such as for scheduled synchronization services, and profitable inelastic traffic, such as for real-time streaming services. Admission control in this setting is cast as a stochastic shortest path problem, with state space derived from (discretization of) the elastic flow's size/deadline and action space corresponding to alternative subsets of admitted inelastic flows: the probabilistic model expresses uncertainty in both the link's available bandwidth and the inelastic flows' offered loads, while the objective function captures both congestion avoidance and the option to specify a desired minimum elastic rate. Its solution is shown to (i) balance the accumulation of instantaneous inelastic reward with the risk of missing the elastic deadline and (ii) exhibit a degree of robustness to link & flow modeling errors that is tunable via choice of the desired minimum elastic rate. Also discussed are state augmentations that befit urgent or non-interruptible inelastic traffic.