Award Abstract #1409367
CSR: Medium: Collaborative Research: Architecture and System Support for Power-Agile Computing

NSF Org:	CNS Division Of Computer and Network Systems
Initial Amendment Date:	July 30, 2014
Latest Amendment Date:	August 18, 2015
Award Number:	1409367
Award Instrument:	Continuing grant
Program Manager:	M. Mimi McClure CNS Division Of Computer and Network Systems CSE Direct For Computer & Info Scie & Enginr
Start Date:	August 1, 2014
End Date:	July 31, 2017 (Estimated)
Awarded Amount to Date:	$282,931.00
Investigator(s):	Geoffrey Challen challen@buffalo.edu (Principal Investigator)
Sponsor:	SUNY at Buffalo 520 Lee Entrance Amherst, NY 14228-2567 (716)645-2634
NSF Program(s):	COMPUTER SYSTEMS
Program Reference Code(s):	7924
Program Element Code(s):	7354

ABSTRACT

Short battery lifetimes are one of the top complaints of today's smartphone users and are limiting the transformative potential of these mobile devices. Hardware components continue to offer new energy-saving features, but today's mobile operating systems are ill-prepared to utilize these new capabilities. This project investigates novel hardware-software coordination mechanisms to make use of existing and next-generation hardware energy-saving features more effectively. The focus is improving "power agility", a measure of a system's ability to adaptively reallocate energy between hardware components to meet the changing needs of running applications, through the introduction of new operating system interfaces and abstractions.

In contrast to previous approaches that attempt to manage energy directly, this project explores the idea of measuring and managing energy inefficiency, defined as the amount of extra energy actually used to perform a task above the minimum required for the task. Allocating inefficiency, rather than energy, addresses many of the problems with previous energy management approaches while still allowing the operating system to prioritize energy usage between tasks. By allocating inefficiency, power-agile operating systems can safely allow tasks to tune hardware components directly using userspace tuning libraries, allowing them to make cross-component tradeoffs in response to changing requirements while pursuing a wide variety of different energy management strategies. The project also integrates education with the research through a new course on power-agile computing at both of the collaborating institutions.

Please report errors in award information by writing to: awardsearch@nsf.gov.