Christian Kästner is an associate professor in the School of Computer Science at Carnegie Mellon University. He received his PhD in 2010 from the University of Magdeburg, Germany, for his work on virtual separation of concerns. For his dissertation he received the prestigious GI Dissertation Award. Kästner develops mechanisms, languages, and tools to implement variability in a disciplined way despite imperfect modularity, to understand feature interactions and interoperability issues, to detect errors, to help with nonmodular changes, and to improve program comprehension in software systems, typically systems with a high amount of variability. Among others, Kästner has developed approaches to parse and type check all compile-time configurations of the Linux kernel in the TypeChef project.




Performance Analysis for Highly-Configurable Systems

Almost every modern software system is highly configurable with dozens or more options to customize behavior for different use cases. Beyond enabling or disabling optional functionality, configuration options often adjust tradeoffs among accuracy, performance, security, and other qualities. However with possible interactions among options and an exponentially exploding configuration space, reasoning about the the impact of configurations is challenging. Which options affect performance or accuracy? Which options interact? What’s the optimal configuration for a given workload? In this talk, I will given an overview of different strategies and challenges to learn performance models from highly-configurable systems by observing their behavior in different configurations, looking at sampling and learning strategies, transfer learning strategies, and strategies that analyze the internals or architecture of the system.