This proposal is to develop a unified approach in operating systems (Oss) for managing energy consumption. The goal is to make energy management completely automatic for all types of hardware components. This approach can be divided into three parts: (1) creating an interface for characterizing hardware components, (2) constructing energy models for supporting the Oss' run-time decisions, and (3) developing techniques to determine components' power states for managing energy.

Using Oss for energy management is still in its infancy. Timeout is a widely used (in most cases the only) solution in real-world Oss. Moreover, users have to set the lengths of the timeout. No operating systems can automatically find the appropriate timeout values based on the component's characteristics or the executed programs.

The first part of this proposal creates an interface for automatically characterizing components' power consumption. This hides the differences among components so that the Oss can manage the energy of all components in a uniform way. The second part of this proposal constructs energy models for estimating energy consumption; the models' coefficients can be automatically determined by using the proposed interface. The third part of this proposal develops techniques for determining components' power states. The OSs automatically select appropriate algorithms and set the algorithms' parameters. The parameters are calculated based on the components' characteristics. The algorithms change components' power states, for example by scaling down the performance of under-utilized components.

The education plan emphasizes system integration and helps students develop the skills to discover, define and solve complex problems. This plan includes three courses: (1) a graduate course on embedded systems, (2) an undergraduate course on computer organization, and (3) a graduate course on computer architectures. Under-represented minorities, disadvantaged students, women and undergraduate students will be recruited to participate in the research and education programs.

Agency
National Science Foundation (NSF)
Institute
Division of Computer and Network Systems (CNS)
Application #
0347466
Program Officer
Krishna Kant
Project Start
Project End
Budget Start
2004-06-01
Budget End
2010-05-31
Support Year
Fiscal Year
2003
Total Cost
$431,775
Indirect Cost
Name
Purdue University
Department
Type
DUNS #
City
West Lafayette
State
IN
Country
United States
Zip Code
47907