Adaptive Isolation for Predictability and Security

This report documents the program and the outcomes of Dagstuhl Seminar 16441 “Adaptive Isolation for Predictability and Security”. Semiconductor technology is at the verge of integrating hundreds of processor cores on a single device. Indeed, affordable multi-processor system-on-achip (MPSoC) technology is becoming available. It is already heavily used for acceleration of applications from domains of graphics, gaming (e.g., GPUs) and high performance computing (e.g., Xeon Phi). The potential of MPSoCs is yet to explode for novel application areas of embedded and cyber-physical systems such as the domains of automotive (e.g., driver assistance systems), industrial automation and avionics where non-functional aspects of program execution must be enforceable. Instead of best-effort and average performance, these real-time applications demand timing predictability and/or security levels specifiable on a per-application basis. Therefore the cross-cutting topics of the seminar were methods for temporal and spatial isolation. These methods were discussed for their capabilities to enforce the above non-functional properties without sacrificing any efficiency or resource utilization. To be able to provide isolation instantaneously, e.g., even for just segments of a program under execution, adaptivity is essential at all hardwareand software layers. Support for adaptivity was the second focal aspect of the seminar. Here, virtualization and new adaptive resource reservation protocols were discussed and analyzed for their capabilities to provide application/job-wise predictable program execution qualities on demand at some costs and overheads. If the overhead can be kept low, there is a chance that adaptive isolation, the title of the seminar, may enable the adoption of MPSoC technology for many new application areas of embedded systems. Seminar October 30–4, 2016 – http://www.dagstuhl.de/16441 1998 ACM Subject Classification C.3 Special-Purpose and Application-Based Systems: Realtime and embedded systems