Despite significant scientific research, systematic performance engineering techniques are still hardly used in industry, as many practitioners rely on ad-hoc performance firefighting. It is still not well understood where more sophisticated performance modeling approaches are appropriate and the maturity of the existing tools and processes can be improved. While there have been several industrial case studies on performance modeling in the last few years, more experience is needed to better understand the constraints in practice and to optimize existing tool-chains.
I gave a talk summarizing six years of performance modeling at ABB. In three projects, different approaches to performance modeling were taken, and experiences on the capabilities and limitations of existing tools were gathered. The talk reports on several lessons learned from these projects, for example the need for more efficient performance modeling and the integration of measurement and modeling tools.