CAREER-EHS: Thermal-Aware Task Scheduling for Embedded Planar and 3D Chip Multiprocessors
University Of Pittsburgh, Pittsburgh PA
Investigators
Abstract
Many mobile embedded systems are designed to be small and compact to favor portability. As the user demand expands for more powerful, versatile, and integrated solutions, the designers endeavor to pack more and more devices into small embedded form factors, thanks to technology advancement. In parallel with this trend, the microprocessor technology has evolved into an era of multicore processor in both 2D and 3D space. As can be foreseen, the marriage of future embedded systems and future microprocessors raises the specter of dramatically increased power density that renders thermal management a key challenge for embedded processors This CAREER project addresses this challenge and seeks innovative solutions to tackle thermal problems for embedded chip multiprocessors (CMPs). The objective of this research is to develop proactive thermal management techniques that prevent the temperature from increasing above the threshold and yet avoid performance throttling. This is in contrast to traditional techniques, which only react to thermal violations, and do so by enforcing performance throttling to cool down the processor. The proposed techniques leverage the natural discrepancies in thermal behavior among different applications, and schedule them among multiple cores to keep the chip temperature within a given budget. The mission of such scheduling is to minimize thermal violations across all cores on-chip, improve performance, and diminish overheating-induced problems such as reduced reliability, low circuit speed, and high leakage power. The education component of this project seeks to train students from different fields and with different backgrounds in a tightly integrated framework. The goal is to establish broad familiarity and preparedness for future professions in the field of embedded systems, micoarchitectures, operating systems and software analysis.
View original record on NSF Award Search →