NGS: Collaborative Research: Adaptive Performance and Power Management for Real-Time Systems

NGS：协作研究：实时系统的自适应性能和电源管理

• 批准号: 0102539
• 负责人: Yann-Hang Lee
• 金额: $ 21.49万
• 依托单位: Arizona State University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2001
• 资助国家: 美国
• 起止时间: 2001-09-15 至 2005-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0102539&HistoricalAwards=false
• 关键词: NGS; Collaborative; Research; Adaptive; Performance

EIA-0102539Yann-Hang LeeArizona State UniversityAdaptive Performance and Power Management for Real-Time SystemsThis research will enable real-time computer systems to monitor and automatically adapt their power usage to specified fault-tolerance and performance requirements, prevailing workload, and current and projected energy/power constraints.Many systems, which are power-constrained, do not always have a strict limit on their power consumption. Instead, they have periods of time during which energy supply is very limited, and should be used sparingly, while at other times, the power available is adequate and the system is allowed to exploit all its resources to deliver maximum performability. While there has been a great deal of research in developing circuits and design techniques to build low-power devices, there is very little reported on techniques to adjust power consumption on-the-fly to adapt to changing levels of workload and energy/power availability. There is the need for an integrated approach, ranging from the hardware operation (using voltage-clock scaling or a sleep mode) to the application level. Our preliminary investigations have shown that such an integrated approach, exploiting the synergy between these various layers, is far more effective than single-layer approaches adopted incrementally.

EIA-0102539Yann-Hang Lee亚利桑那州立大学实时系统的自适应性能和电源管理这项研究将使实时计算机系统能够监控并自动调整其电源使用，以适应指定的容错和性能要求、当前工作负载以及当前和预计的能源/功率限制。许多功率受限的系统并不总是对其功耗有严格的限制。 相反，在某些时间段内，能源供应非常有限，应谨慎使用，而在其他时间，可用电力充足，系统可以利用其所有资源来提供最大的性能。 尽管在开发电路和设计技术来构建低功耗设备方面已经进行了大量研究，但关于动态调整功耗以适应不断变化的工作负载和能量/功率可用性水平的技术的报道却很少。 需要一种集成方法，范围从硬件操作（使用电压时钟缩放或睡眠模式）到应用程序级别。 我们的初步调查表明，这种利用不同层之间协同作用的集成方法比逐步采用的单层方法要有效得多。