Year

1996

Degree Name

Doctor of Philosophy

Abstract

Control systems in exchanges operating in the multi-service telecommunication environment will be required to support a wide variety of services. Call processing resource and quality of service requirements may vary greatly for different services and call processing complexity is expected to increase as these services become more sophisticated. This will increase the importance of effective performance evaluation tools to dimension and optimise control systems in response to changing call loads and call mixes.

This thesis is concerned with analytical performance evaluation techniques suitable for hierarchically distributed control systems which are required to handle disparate mixes of calls. These techniques are based on low and high-level queueing models and are used for investigations of throughput and cost optimisations.

In the first part of the thesis we review operation and technology issues relevant to distributed control systems. These issues include optimal allocation of call control functions amongst processors and call load balancing. Subsequently, w e examine how these issues can be tackled using existing queueing models. We find that existing methods must be augmented to be useful for performance evaluation of distributed control systems handling different mixes of calls.

We subsequently extend the available queueing network methods for performance evaluation and optimisation of hierarchical control system configurations. Using these methods we examine function allocation as a means of optimising throughput in existing control systems configurations and as a means of producing optimal control systems configurations given particular call mix and call performance requirements. We also examine how the queueing models can be used for overall cost optimisation of control systems.

Based on our investigations of function allocation and cost optimisation we examine the application of high-level queueing models (based on M/G/l queueing networks) and low-level queueing models (based on networks of M/G/l queues with feedback and priorities). We consider this by investigating their relative accuracy and computational complexity.

The work performed in this thesis shows that: 1. High-level queueing models of call processing produce accurate results for a wide range of applications making them suitable for performance evaluation and optimisation during system design. 2. Function allocation can be used as an effective means of optimising throughputs of existing distributed control systems. 3. Function allocation is also an important variable for overall cost optimisation of control systems required to operate in the multi-service environment.

Share

COinS