Analysing system behaviour on different time scales
Suzanne P. Evans · 1996
Abstract Events in an integrated services telecommunications network occur on multiple time scales and careful analysis of time dependent behaviour is important to derive appropriate system descriptions at different levels. This paper analyses time dependent behaviour expressed in terms of the eigenstructure of a Markov generator. Identifying terms that are dominant on different time scales is not always straightforward and we discuss a number of technical conditions required to guarantee a separation of time scales. These conditions are shown to have implications for the appropriate representation of fast and slow time scale dynamics and for the sensitivity of these representations to errors in the model parameters.