Research Spotlight Archive
Title: Energy and Performance-Aware Run-Time Management of Manycore Systems
Background: Single-chip multicore systems have become increasingly attractive in recent years because of their potential to provide higher throughput per watt in comparison to single-core systems. It has not been possible, however, to achieve the projected ideal peak performance due to high power densities, prohibitive cooling costs, thermal gradients, suboptimal resource utilization due to dynamically varying workloads, and limited memory bandwidth. As we enter the manycore era, where a single chip has dozens to hundreds of cores, handling the performance and power challenges at design time is no longer possible because of the design complexity and time-to-market restrictions. Therefore, capability of analyzing and optimizing performance, energy, and temperature at run time is crucial for maximizing the execution efficiency.
Description: The main objective of this project is to develop a suite of run-time management techniques for manycore systems by jointly exploring key contributors to system performance and power: reconfigurable network architectures, workload scheduling policies, and microarchitectural resource management. These techniques will be designed using an integrated hardware-software approach to dynamically monitor system behavior and to enforce intelligent decisions for improving performance, energy efficiency, and thermal efficiency.
In addition to the 48-core Single-Chip Cloud computer provided by Intel for this project, we utilize a set of simulation and emulation tools to aid in the analysis and design.
Recent Publications: http://people.bu.edu/joshi/Site/ICSG__Publications.html