Developing novel statistical bandwidths for communication networks with incomplete information

Abstract

In this paper, the concept of statistical bandwidth of multi-access systems are studied and extended to the case of unknown statistical descriptors. The results can improve the statistical characterization of the tail distribution of aggregated load presented to a multi-access system which is traditionally based on the logarithmic moment generation function (LMGF)[1]. In the paper, an extended moment generating function is introduced for calculating the statistical bandwidth and as a result a novel admission algorithm is presented. To further maximize the admitted load into the multi-access system the free parameter of the extended statistical bandwidth is optimized based on the geometrical optimization of polygonal surfaces. In this way, the system utilization can be near-optimal. © Springer-Verlag Berlin Heidelberg 2005.