Algorithmic Support for Temperature Aware Computing and Networking

点击翻译按钮获取中文摘要

Recently, it has been recognized that power management to control temperature is vitally important. In fact, in May Intel abruptly announced that it had scrapped the development of two new computer chips (code named Tejas and Jayhawk) for desktops/servers in order to rush to the marketplace a more efficient chip technology more than a year ahead of schedule. Analysts said the move showed how eager the world's largest chip maker was to cut back on the heat its chips generate. Intel's method of cranking up chip speed was beginning to require expensive and noisy cooling systems for computers. There is an extensive literature on power management in computing devices, and there is an extensive literature on power management in ad-hoc and sensor networks. Yet almost all of this literature focuses on power management to conserve energy, and not on power management to reduce temperature. Temperature and energy are different physical entities with quite different properties. Bluntly, if the processor in your mobile device exceeds its energy bound, your battery is exhausted, but if your mobile device exceeds it thermal threshold, your processor dies.The need to manage power to reduce temperature will become increasingly important as device power consumption is growing exponentially, and as ad-hoc and sensor networks become more prevalent. We propose to reexamine power management problems from a temperature perspective to try to better understand the difference between managing power to conserve energy and managing power to minimize temperature. We plan to start with speed scaling problems within devices and with routing problems in ad-hoc networks. It is anticipated that this project will produce novel optimization techniques and analytical results that will be useful in the design of temperature-efficient protocols for large-scale ad-hoc and sensor networks. The work will be a joint inter-disciplinary effort, involving investigators with expertise in on-line algorithm design and in network protocol design and development. It is further anticipated that the project will provide students with a unique opportunity to investigate and experiment with new ways of thinking about important design parameters for this emerging class of networks.