UT ECE professor Vijay Janapa Reddi has been awarded awarded a National Science Foundation (NSF) grant to investigate Cross-Layer Solutions for Sustainable and Reliable Computing Systems with Co-PI Dr. Lizy John.
A critical research challenge facing the semiconductor industry is designing future high performance and energy efficient systems while satisfying historical standards for reliability and lower costs. The primary source of this challenge is the device and circuit parameter variability, resulting from the manufacturing process as well as the wear and tear of daily system operation. As technology scales to increasingly smaller transistor geometries, the adverse impact of variations on system metrics will increase.
The goal of this research is to develop a resilient software-assisted, hardware-guaranteed machine organization that optimizes reliability, energy and performance, in a coordinated manner. In this system, the software is continuously monitoring system execution and adapting the hardware’s configuration based on specific QoS feedback. The feedback-driven system relies on the hardware for an immediate reaction (albeit suboptimally) to reliability emergencies, and utilizes the software’s global knowledge of activities across the entire system to carefully optimize repeated occurrences of such hardware emergencies. The research will also result in new runtime system algorithms and techniques corresponding to reliability on emerging multicore and manycore CPUs and GPUs architectures, respectively, running Internet type-of cloud computing workloads.