Modern advances in software technologies for today’s application development have allowed for developers to concentrate less on issues, such as performance and resource management, and instead spend more time on developing the application functionality such that the time to market is reduced. Consequently, performance analysis and optimization become more difficult and create a need for advanced performance tools that should provide a clear report of the application in terms of its performance and allow a fast interpretation of these results. This work investigates typical performance problems of today’s applications and offers approaches on how to automatically detect them. Using statistical methods like regression and correlation analysis, investigation of measured values is performed in order to detect performance deviations that possibly occurred in an application under test. The proposed approach has been implemented as a new Reporter component into an open source performance testing tool PerfCake, developed by QE engineers from Red Hat Czech s.r.o. The developed component is capable of detecting and reporting possible issues and their probability. A unified report from all pre-specified measurements is created in such a way that all detected performance issues are immediately visible. The aim is to improve an end-user experience and usability when reading the report from performance testing.
Unified Reporting for Performance Testing
Faculty of Information Technology
Date of Completion