TECoSA Seminar – Deterministic Reactive Software for Embedded, Edge, and Cloud Systems
June 3, 15:00 – 16:00
We aim to bring you a TECoSA Seminar at kl.15 on the first Thursday of each month. This Spring they will be on-line, and all are welcome to join. Each invited speaker will talk for about 40 minutes, followed by a panel discussion coordinated by TECoSA members.
The eighth speaker in our series is Dr Martin Lohstroh of UC Berkeley, Electrical Engineering and Computer Sciences. Marten is a postdoctoral researcher advised by Professor Edward A. Lee, who was also his doctoral adviser. Since obtaining his PhD from Berkeley, he has remained focused on disproving the dogma that determinism is fundamentally at odds with high performance, and making the case that time is a valuable and under leveraged abstraction in computing. In 2018, he was a recipient of the Systems and Software Modeling (SoSyM) Best Paper Award. Marten also holds a BSc in Computer Science and MSc in Grid Computing, both from the University of Amsterdam.
Please email email@example.com to register!
Deterministic Reactive Software for Embedded, Edge, and Cloud Systems
ABSTRACT: Determinism is notoriously hard to achieve in concurrent systems, and this poses serious challenges to our ability to build confidence in the correctness of their implementation. The growing pervasiveness of networked computing and a trend towards integrating computationally demanding artificial intelligence components into real-time cyber-physical systems (think of robotics, autonomous vehicles, etc.) add urgency to this problem. This talk is about a solution in the form of a concurrent programming model called reactors (actors, revisited), that delivers determinism by default. The execution of reactors is underpinned by a temporal model that incorporates a multiplicity of timelines—some logical, relating to discrete synchronous-reactive ticks, and some physical, relating to the passage of wall-clock time. The relationship that reactors establish between events across timelines allows for the construction of programs that react predictably to unpredictable external events. Moreover, determinism in reactor systems can be maintained in a distributed setting under quantifiable assumptions. We discuss this work through the lens of Lingua Franca (LF), a polyglot reactor-oriented coordination language developed in a collaboration between UC Berkeley, UT Dallas, TU Dresden, and Kiel University. Distributed LF programs may span deeply embedded, edge, and cloud systems.