Between rising energy costs for large data centers and the need to extend the life of mobile devices, when it comes to designing a computing system, energy consumption is now a top priority. Microsoft wants to reduce the energy costs of idling computers as well as better manage peak computing energy demands. To this end Microsoft will dole out half a million dollars to four university research groups, all of whom have been working on very different aspects of “sustainable computing.”
The grants are out the company’s sustainable computing program. Microsoft says data centers account for 1.5 percent of the electricity consumed in the U.S. and estimate that that number will rise to 4.5 percent in just five years. The need for new, power-conscious design is critical to the world’s information infrastructure, which is growing at a rate far greater than the world’s energy capacity.
- Control-Theoretic Power and Performance Management for Green Data Centers; Xiaorui Wang, University of Tennessee. Flipping computing priorities around, this project makes energy the top priority and adjusts hardware and software performance according to power constraints. The system’s power/performance state is optimized with no regard for application-level performance, meaning an operator can’t simultaneously guarantee application performance and power consumption. To deal with this, the project is creating multi-input-multi-output algorithms to create an energy and performance management system.
- Building a Building-scale Power Analysis Infrastructure; Phil Levis, Christos Kozyrakis, Nick McKeown, Stanford University. Managing a single data center is one thing, but tracking the energy use of an entire building of servers, infrastructure nodes and individual computers is a very different problem. This group wants to deploy a dense set of sensors inside Gates Hall, the computer science building at Stanford, to do the monitoring. The goal is to create a comprehensive, pan-device energy consumption model that can correlate use to energy for a variety of computing needs.
- A Synergistic Approach to Adaptive Power Management; David Brooks, et al., Harvard University. Just as hardware is getting smarter about energy use, software needs to follow suit. This requires creating software that is aware of the dynamic environment in which it is running. By employing an internal energy monitoring system, software can gauge energy use and react by modifying its own performance.
- Simulating Low Power x86 Architectures with Sooner, a Phoenix-based Simulation Framework; Ronald Barnes, University of Oklahoma. This project’s goal is to create a tool to measure low-power micro architectural processes. Using the Microsoft Phoenix tool, the group is working to create a simulation framework in which to run tests within.
While these four research endeavors could definitely yield large scale power savings for data center operators and better power diagnostics we wonder what it would take to just help home and office PC users better employ the built in power saving options Windows has available. That is some low-hanging green computing fruit but it’s more a distributed behavioral problem than a technical engineering challenge.