Paging Performance in GSM Networks: Analysis and Simulation

摘要

In this paper the performance of the GSM paging process is studied using a combination of analysis and simulation. The system is modelled as a discrete-time Markov chain, representing a single server queue with a finite buffer. The complexity level exceeds that of a classical birth-death process because the queue service discipline allows type-dependent batch departures during each deterministic service interval, where these departing pages may also return in the event of failure. Paging message arrivals are thus not a Poisson process but we show that they can be approximated by one under realistic traffic and blocking levels (up to system utilisation of ～0.9). We analytically derive the queue occupancy probability density function, and draw attention to the effects of quantisation error introduced through the model's time discretisation. The distribution of paging message system time is obtained from a detailed event driven simulation, which is also used to validate the queue occupancy model and region where a Poisson arrival process is applicable. Queueing performance is also found to depend on the percentage of first-try TMSI's which are lost due to bad radio conditions (defined as the probability of TMSI error). The combination of analysis and simulation provides a practically significant solution to understanding paging performance in GSM networks.