Model-Based Performance Testing Using the MBPeT Tool

This document describes a model-based performance testing tool, called MBPeT, used for generating synthetic workload from probabilistic models. The purpose of the tool is two fold: (a) to simplify the creation and updating of test specifications by using abstract models for the user profiles and (b) to create, based on a certain level of randomness introduced by the probabilistic models, synthetic workload which is closer to the real workload as compared to the workload generated from static scripts. MBPeT has a distributed architecture, where one master node controls multiple remote slave nodes. Each slave node is capable of generating synthetic workload based on the specified load profile. Besides measuring different key performance indicators of the system under test, the tool monitors permanently the resource utilization of its slave nodes and (when possible) the one of the system under test. At the end of each test run, the measurements are aggregated by the master into a detailed test report. This report describes the implementation details behind MBPeT and provides several experiments highlighting the capabilities of the tool.