Performance analysis of ATM switch with call admission control

Abstract

Asynchronous transfer mode (ATM) networks will provide multimedia services with diversified traffic flow characteristics and quality requirements. A prime instrument for controlling congestion in the network is admission control, which limits calls and guarantees a quality of service (QOS) determined by delay and loss probability in the switching nodes. Though there have been many works on ATM multiplexers with connection admission control (CAC), studies of performance analysis or proposals of effective CAC schemes in ATM switch are rare. The main objective of this dissertation is the performance study of ATM switches with CAC. First, we investigate the performance of an input queueing ATM switch under the condition that CAC is applied to the switch. For the analysis, we assume that calls are homogeneous and their lengths are geometrically distributed. In the switch with CAC, each input and output port of the switch has a finite number of connected calls, respectively. Using this assumption, we model the arrival process as a discrete-time MMBP driven by call states existing in input and output ports. We obtain the average input queue length and cell loss probability through Markov chain analyses applied to the virtual and input queue. The dimension of the chains is proportional to the maximum number of calls being able to accept in each input and output port, but not to the switch size. By numerical analysis, we investigate the effect of CAC, the number of calls, burst length and the buffer size in input queues. The results illustrate that when the CAC is adopted, the effect of burst length is large. Also we propose a call-based distribution scheme which can be applied to the switch architectures having the cell distribution scheme. Applying this scheme to a dual plane switch architecture, we show through analysis and simulation that the performance of the proposed scheme is superior to that of the classical random distribution. Second, we analyze an output que...