On autonomic QoS provisioning for Voice Over IP services

In this paper we present a framework for autonomic QoS provisioning for VoIP services. The central role is held by a Measurement Based Admission Control algorithm which incorporates three innovations. First, the mathematical underpining obeys a non-parametric approach, removing the dependency on a priori assumed characteristics of the underlying stochastic process, i.e. those of the work arrival process. Thus, we tackle a major issue of common parametric MBAC. The second and the third enhancement is embodied in the combination of a closed-loop control based on perceived QoS. Typically, MBAC algorithms do not validate there decision and thats why many algorithms miss QoS targets due to non-stationaries in work arrival processes. As a metric for performance evaluation we devised a new approach based on the Emodel, an ITU-T standard for quantifying QoS assessment based on human perception. The contribution is the mathematical framework for both, the non-parametric MBAC and its closed-loop control but also an analysis based on simulations.