Bit Loss Probability versus Loss Probability in the Admission Control Process of Non-Elastic Traffic Sources

Admission control is one of the key functions performed by the management of services and network resources in order to guarantee the QoS requirements of different types of traffic flows. In the case of non-elastic traffic streams, the admission control is typically based on a formula derived by assuming that the traffic aggregate may be approximated by a Gaussian distribution and considering the loss probability, i.e. the buffer overflow probability, as the QoS parameter. In this paper we consider an alternative QoS parameter that is the bit loss probability that is the long term proportion of lost bits, which is more related to the current measurements (e.g. throughput or packet loss rate) than the loss probability. Therefore, we derive an appropriate computational formula for the effective bandwidth which, as numerical results show, is more accurate and, consequently, assures a more efficient admission control mechanism.