How to build robust call admission control based on on-line measurements

In this paper we are concerned with the robustness of measurement-based call admission control (CAC) procedures regarding the guarantees on quality of service (QoS) parameters. In the case of any measurement based CAC the probability of a buffer overflow is generally caused by the combination of two effects. One comes from the statistical behaviour of the sources, while the other one is the effect of measurement errors that allow too many sources admitted into the system which in turn results in QoS violation. If we want to provide guaranteed QoS we have to control these two effects together. The algorithms available in the literature calculate the loss probability taking into account only the first affect and none of them attempts to give a bound on the probability of a bad CAC decision due to measurement uncertainty. The appealing feature of our method is that we can take into account both of the above affects, which in turn results in robust call admission control with real guarantees on QoS parameters. The basic idea of our procedure can be reused in case of any measurement-based CAC, however in this paper we illustrate it for the Mosquito algorithm (Crosby et al. 1997)