Multi-Class Measurement Based Admission Control for a QoS Framework with Dynamic Resource Management

The advent of a Premium Service (PS) featuring Internet is still pending, mainly due its difficult deployment. But do we really need Quality of Service like PS? Likely no. We are therefore investigating an alternative QoS model, which is more flexible to deploy at the expense of looser QoS guarantees. We call it a Better-than-Best-Effort (BBE) service. What distinguishes our approach from traditional ones is dynamic resource management based on measurements and ruled by Perceptual QoS. What yet has been missing was Admission Control (AC) and in this paper we present our latest advance. Based on the existence of measurements taken by the queueing module, we developed a Measurement Based AC algorithm. Design goal was simplicity and general applicability in terms of independence from statistical assumptions. Likely the most interesting finding is that even a very simple design proofed to be reasonably effective for our BBE service, mainly due to a cross-layer design, i.e. cooperation between dynamic resource management on queuing level and AC. This is the conclusion a comprehensive simulative performance evaluating