The Effects of Traffic Load Variation on Measurement Accuracy

The Effects of Traffic Load Variation on Measurement Accuracy By D. W. HILL (Manuscript received June 10, 1976) A method is developed to quantify the effect of day-to-day variation in offered load on the accuracy of functions of traffic measurements. The method is applicable to any smooth function of the standard trunk-measurements--i.e., peg count, overflow, and usage. As an example, the accuracy of the trunks-required estimator for probabilityengineered, full-access trunk groups is approximated. A sensitivity analysis shows that the major contributor to the variance of the estimator is day-to-day variation. I. INTRODUCTION Most of the traffic-engineering procedures in use in the Bell System require data collected over periods of up to several days. It has long been recognized that the daily offered loads estimated from trunk-group data show considerable variability even for data taken during the same hour of successive days.1,2 This variability can cause significant differences between the observed blocking and the objective grade of service. It also can induce large fluctuations in the estimation of network requirements and must be accounted for in the traffic-engineering procedures. (For details on a model for this day-to-day load variation, see the Appendix.) T w o important applications that can be affected by day-to-day load variation are trunk servicing and trunk forecasting. The former is the use of traffic measurements to determine when trunk groups are significantly overloaded or underloaded.