Efficient admission control for EDF scheduler with statistical QoS guarantee

Abstract

We study the admission control problem at an earliest deadline first (EDF) link scheduler where multiple quality of service (QoS) requirements are supported statistically. It is shown that, if the aggregate QoS requirement of the aggregate traffic is guaranteed, the individual QoS requirements can also be guaranteed using an optimal discarding policy when the traffic flows are equally demanding. Therefore the problem is reduced to one with single traffic stream and a single QoS requirement. We adopt the stationary Gaussian approximation method and present a simple formula, referred to as the weighted Gaussian method, to estimate the minimum bandwidth requirement. Efficient admission control algorithms can be readily constructed based on the weighted Gaussian method. Numerical results are presented to show its advantages and limitations.