Performance statistics of a time sharing network at a small university

Abstract

The purpose of the study was to characterize the input traffic on the interactive terminal facility (ITF) at Jackson State University by mathematical distributions. Five traffic measures were extracted from a week's job accounting records: arrival time (time of day of log on), interarrival time (time between successive arrivals), active central processing unit (CPU) time (user's time slice), inactive CPU time (swapping time), and connect time (duration of a user session). To facilitate analysis, four user groups were formed. An analysis of variance established that the group category largely determined the volume of arrivals. Since a normal distribution could not be fitted to the data, it was concluded that the pattern of daily arrivals was best treated as an empirical distribution. Using the method of moments to estimate parameters, it was found that interarrival times could best be characterized by a gamma distribution, while connect, inactive CPU, and active CPU times approximated exponential distributions.