A framework for real-time measurement-based connection admission control in multi-service networks

Abstract

We introduce a new framework for measurement-based connection admission control (CAC) for multi-service networks. The framework encompasses function modules that allow customization for varying degrees of complexity and efficiency. In all methods, when a new connection arrives, the CAC makes a decision to accept or reject the connection according to a novel procedure for free bandwidth evaluation. This procedure is based on real-time measurements and an adaptive feedback mechanism. In addition, multiple time-scale histograms keep track of the aggregate traffic per link. Simulation results show that a utilization level of over 70% was achieved whilst still meeting the quality of service guarantee. Real variable bit rate traffic was used as video traces.