Approximate method for polling systems with time-limited-based polling tables

Time-limited based polling systems with polling tables are encountered frequently in operating systems schedulers such as the fair share scheduler. The analysis of such systems are usually carried out by simulation. We present a simple approximation for determining the mean waiting times in such polling systems. This approximation can be used by system designers to carry out a first cut design to reduce the set of alternative designs and then use simulation to assess a few good designs. The method transforms the M table polling system with N distinct stations, (N⩽M), to an equivalent M pseudostations cyclic polling system. It then uses the known K-limited polling results by approximating K from the time limit and the mean service time of each station. Polling systems are a class of multiqueueing systems attended to a single server. Such systems are encountered very frequently in communications and computer, traffic signal, and also manufacturing systems. For computer systems in a multiprogramming environment we would like to keep the CPU busy all the time (a 100% utilization would be optimal)