Monte Carlo based user activity simulation for software performance evaluation

The methodology of estimation of reliability characteristics of computer networks proposed earlier is supplemented with a new technique for estimating the software performance metrics. The proposed technique is based on the Monte-Carlo simulation method. To apply simulation in the task of analyzing software metrics, the following problems are solved: ($i$) a model of the software component is defined — a program that implements a three-tier architecture and executes user queries or commands; ($ii$) a model of user's activity is defined — a nonlinear discrete-time stochastic system based on the autoregressive model with switching using classification of process states; ($iii$) a simulation program is described that implements the model of user requests, generates the user requests flow, and collects statistics; and ($i\nu$) numerical examples illustrating the method's applicability are provided. The concluding remarks contain recommendations concerning further usage of the presented method.