Digital twin based timing bottleneck analysis in large-scale component-based software

To obtain high productivity and accuracy, ASML’s Twinscan machines expose wafers in a single smooth movement. For this purpose such scanners contain a lot of control software correcting for nanometre physical disturbances. This control software is deployed in hundreds of software components operating in real-time. As these components use shared (computational, communication and memory) resources of the execution platform, a slight delay by one can influence the timing of another component. This may result in components that miss deadlines, possibly disturbing the smoothness of the exposure.

To analyse run-time behaviour of component-based software systems, a digital twin approach based on Message Sequence Charts was developed. Through systematic software instrumentation, these digital twins are inferred fully automatically from execution traces of running Twinscan machines. They capture the execution of the component-based software system in an intuitive way and support timing bottlenecks to be pinpointed in a very efficient manner. The approach is warmly received by ASML developers and architects and is currently being incorporated in ASML’s development process.

Read more about the project in this recent article in Artemis Magazine (page 22)

 

About the speaker

Jeroen Voeten received his MSc in Computer Science in 1991 and his PhD in Electrical Engineering at the TUE in 1997. He works as a full professor at the Eindhoven University of Technology and is the scientific director of CEETSe, the TUE center stimulating cross-disciplinary research in cyber-physical systems. Voeten has fifteen years of experience in applied research as a senior scientist at TNO and scientific advisor of TNO-ESI. His professional passion is to improve industrial systems and design processes, by working on the borders between academic research and industrial innovation, by stimulating cross-disciplinary collaboration, and by teaching systems thinking. His research expertise is performance engineering, including the areas of (stochastic) performance analysis, design-space exploration, scheduling and predictable synthesis. He developed and supervised several research programs that led to various industrial innovations and authored over one hundred journal and conference publications.