Integration of performability-model extraction and performability prediction in continuous integration/continuous delivery

Predicting the performance and dependability, or performability for short, of software systems is important to ensure they can fulfil requirements like a certain latency or throughput. While the prediction is often used in the design phase of systems, it can also help during operations and in particular to ensure degrading hardware does not lead to the non-fulfilment of the requirements. To allow for the continuous simulation and validation of the fulfilment of performability requirements of a system, we present an approach that extends and integrates the performance extraction tool Performance Model eXtractor (PMX) into a continuous integration and delivery pipeline. The model extraction uses traces and metrics from real systems to generate a Palladio Component Model (PCM) that is subsequently simulated and the results are presented in the pipeline. This enables the software architects and system operators to predict the system behaviour and react accordingly if requirements can no longer be fulfilled.