Perturbation Analysis of a Token Bus Protocol for Network Performance Management

This paper presents an application of perturbation analysis (PA) to performance management of multiple-access computer networks. It is shown how the PA technique enables a discrete-event simulation program to calculate the network performance stimulated by perturbation in priority timer settings of a linear token passing protocol in concurrence with simulation or on-line operation of the network with nominal parameter settings. By feeding this information to the stochastic optimization program(s) (that could be resident in the network management system), the priority timer settings are updated to improve the network performance. This procedure for performance management offers a step forward to bridging the gap between management standards and users' demands for efficient network operation since most standards such as ISO OSI and IEEE address only the network management architecture and interfaces. An on-line performance management procedure that uses the concepts of perturbation analysis, stochastic optimization and learning automaton is proposed for future research.