Capacity Planning with a Stochastic Workflow Model

Abstract

This paper presents a new model for capacity planning which captures the influence of queueing behavior on resource utilization. The model is compared to a similar deterministic model which captures only the workload distribution effects, and assumes that full utilization of the bottleneck resource can be achieved. It is shown that the models are closely related, and serve to complement one another in providing understanding of the capacity estimation problem. Because the stochastic model is nearly as easy to use and provides more realistic results, it is favored over the deterministic model for actual use in planning.