Software is fulfilling more and more complex tasks. As a consequence, the software is getting more complex as well. Model-driven software development helps developers to predict quality aspects like performance or maintainability before the implementation. Palladio is a tool that supports model-driven software development, by creating different models, which together become the Palladio Component Model. It can predict the quality attributes of the software by using analytical or simulation-based solvers.

As shown in previous work, CPU simulators can predict the performance of parallel software. In combination with the code generated from the Palladio Component Model, the simulator can predict the performance before the implementation. As of now, however, this is time-consuming and consists of several manual steps.

This thesis describes steps to automate the manual tasks of the performance prediction and the integration into Palladio. It discusses the problems that arise from such an automation and why I decided not to implement this approach.

The results are a detailed description of the problems, the main one being that simulators only supporting up to Java 7. In addition, the problem of the software having external dependencies was found and solved for the simulators. Finally, I list different ways to solve the problems to get an automated performance prediction in Palladio.