Abstract
|
An important performance aspect of modern switching systems and signalling networks is the response time to user requests. The overall system consists of nodes communicating by exchanging messages. In each node incoming messages pass a series of service phases with possibly different priorities. Therefore a node can approximately be modelled as an M/G/1 priority system with feed-back. The mean transfer times of different messages are a useful measure to describe the system behaviour. In this paper, a generic model is presented including batch arrivals, random branches, and forks. Furthermore different preemption and queueing disciplines can be defined. The model is evaluated with regard to phase utilization and mean transfer time through a single phase or a chain of phases. The analysis uses an efficient and easy to implement algorithm based on the method of moments. Interesting effects occur if the system is saturated. In this case the offered load will become greater than 1 and only higher priority phases are passed in finite transfer times.
|