Date of Award
Doctor of Philosophy (PhD)
Parallel processing is an important way to satisfy the increasingly demanding computational needs of modern real-time and cyber-physical systems, but existing parallel computing technologies primarily emphasize high-throughput and average-case performance metrics, which are largely unsuitable for direct application to real-time, safety-critical contexts. This work contrasts two concurrency platforms designed to achieve predictable worst case parallel performance for soft real-time workloads with millisecond periods and higher. One of these is then the basis for the CyberMech platform, which enables parallel real-time computing for a novel yet representative application called Real-Time Hybrid Simulation (RTHS). RTHS combines demanding parallel real-time computation with real-time simulation and control in an earthquake engineering laboratory environment, and results concerning RTHS characterize a reasonably comprehensive survey of parallel real-time computing in the static context, where the size, shape, timing constraints, and computational requirements of workloads are fixed prior to system runtime. Collectively, these contributions constitute the first published implementations and evaluations of general-purpose concurrency platforms for real-time and cyber-physical systems, explore two fundamentally different design spaces for such systems, and successfully demonstrate the utility and tradeoffs of parallel computing for statically determined real-time and cyber-physical systems.
Christopher Kunal D. Gill Agrawal
James H. Anderson, Roger Chamberlain, I-Ting Angelina Lee, Arun Prakash,