Real-time Applications in a CRMA Network: a Performance Analysis

Abstract

This paper analyzes bandwidth allocation schemes for managing real-time applications (Variable Bit Rate video) in a CRMA network. A methodology to compute the Quality of Service (QoS) experienced by Variable Bit Rate ( VBR ) video sources is proposed. As VBR video applications only tolerate extremely low packet loss rates (< 10 -8 ), we need a computational approach to estimate very low tail probabilities. Studying the QoS with a simulation technique is not feasible, because computational costs make it almost impossible to estimate tail distribution probabilities lower than 10 -2 - 10 -3 . Therefore, to achieve this target, we propose a model which represents a CRMA network's worst case behavior (i.e., a scenario with maximum network congestion), and which can be solved analytically. By solving this model for different bandwidth allocation schemes we can obtain the bounds on the QoS experienced by VBR video users.