Feedback Techniques for Integrating Power Management and Quality of Service in Servers

对电力问题的日益关注和最近提高计算机系统能源效率和散热的方法并没有立即转化为丰富的、适应性强的、对电力敏感的应用。这种有效性的缺乏在功率感知实时计算中尤为明显，在这种计算中，弥合硬件和软件之间差距的努力通常仅限于开发新的调度器，这些调度器通过调制相对高级的硬件控制（如电压缩放和频率缩放）来运行。这种方法的问题是，硬件控制在一个太粗糙的层次上操作，在动态的、建模不良的应用程序场景中是无效的。该项目开发了一种基于反馈的控制理论方法，在运行时优化和平衡功耗和实时性能。这种方法不仅集成了具有实时和服务质量（QoS）约束的硬件级电源管理技术，而且还将一系列不同粒度的硬件电源管理技术集成到一个统一的框架中，从而允许更好的任务级电源管理控制。这使得操作系统能够更好地控制功率/性能权衡，并允许应用程序在每个任务的基础上动态调整系统和芯片的节能配置，以满足特定的性能要求，同时降低功耗。借助该项目提供的新功能，可以开发更丰富的功耗感知应用程序，从而在下一代实时环境中提供有价值的用户体验。该项目特别关注web服务器场的需求，其中能源和冷却成本很高，并且web服务提供不同级别的QoS，从而激发平衡性能，QoS需求和能源成本的功耗感知框架。

Increased awareness of power concerns and recent approaches to improveenergy efficiency and heat dissipation in computer systems have notimmediately translated into rich, adaptable, power-aware applications.This lack of effectiveness is particularly evident in power-awarereal-time computing, where efforts to bridge the gap between hardwareand software have generally been limited to developing new schedulersthat operate by modulating a relatively high-level hardware controlsuch as voltage scaling and frequency scaling. The problem with thisapproach is that the hardware controls operate at too gross a level tobe effective in dynamic, poorly-modeled application scenarios.This project develops a feedback-based, control-theoretic approach tooptimize and balance, at runtime, power consumption and real-timeperformance. This approach not only integrates hardware-levelpower-management techniques with real-time and quality-of-service(QoS) constraints, but also integrates a range of hardwarepower-management techniques at different granularities into a unifiedframework that allows better task-level control of power management.This enables finer control of power/performance tradeoffs by theoperating system, and permits applications to dynamically adjust thesystem's and chip's power-saving configuration on a per-task basis tomeet specific performance requirements while reducing powerconsumption. With the new functionality provided by this project, aricher class of power-aware applications can be developed to providevaluable user experiences in next-generation, real-time environments.The project specifically focuses the requirements of web serverfarms where energy and cooling costs are substantial and where the webservices provide differentiated levels of QoS, thus motivating apower-aware framework that balances performance, QoS requirements, andenergy costs.