Connection admission control - closing the loop

Abstract

To date, many connection admission control (CAC) schemes have been proposed. These schemes can be generally classified into two categories: measurement-based CAC and traffic descriptor-based CAC. Both categories adopt an open-loop architecture where traffic parameters are obtained from either on-line measurements or user-specifications and then input into a chosen traffic and network model to make connection admission decisions. No performance feedback is provided to the CAC scheme. However, no traffic model can be claimed to be accurate for all traffic sources (e.g. voice traffic, data traffic, video traffic) under all network scenarios (e.g. heavy traffic scenario or light traffic scenario). Modelling errors and traffic parameter errors are inevitable. An open-loop CAC scheme lacks the ability to account for modelling errors and adapt to changing network conditions to achieve an optimum performance. In this paper, we proposed a closed-loop architecture for connection admission control. The implementation of the closed-loop architecture in designing a closed-loop CAC scheme is illustrated. Simulation results indicate that the closed-loop architecture is able to overcome the inherent drawbacks of an open-loop CAC scheme to achieve better performance.