CSR: Small: Managing Multicore Energy for Emerging Applications and Devices

多核系统芯片已经渗透到数字市场的每一个细分市场，从服务器和超级计算机，到台式机和智能手机/移动设备。不断涌现的创新应用程序将用户通过移动设备的交互与高端计算和大规模存储的能力相结合。这些应用程序倾向于展示动态和多样化的执行行为，需要密切跟踪以管理资源消耗。固有的应用程序行为变化和资源需求，再加上由于共享资源竞争而导致的复杂交互，对高效、可靠和可持续的计算和信息技术进步提出了新的挑战。本项目将开发基本的系统机制，以管理基于多核的系统上的单个核、内存和功率/能源消耗，并探索使用这些机制的资源分配策略。一个新的在线多核功率模型将允许在细粒度水平上跟踪功率和能量，将现有的基于硬件的功率测量与基于事件的建模和共享资源的功率分配相结合。通过粗粒度测量触发的在线模型重新校准，将提高精度。将开发一种新的操作系统机制，Power Containers，以允许应用程序定义的资源跟踪边界，打破传统的线程和进程边界资源跟踪最先进的操作系统。将探索进一步的优化，以降低资源跟踪的开销，允许在线电源记账扩展到较大的核心和资源计数。最后，将开发新的技术，使电源容器能够以应用程序定义的方式控制和隔离电源使用。该项目将增强基于多核的计算机系统的电源保护、能效和可靠性。开发的技术可广泛应用于数据中心、桌面和移动设备中的新兴动态应用。特别是，PowerContainers将在高端服务器以及多核和多加速器智能手机中识别和缓解(无论是恶意的还是无意的)电源异常(执行会导致异常高的功耗)。该项目的成功将有助于世界不断发展的数字经济和社会的长期可持续性。

Multicore systems-on-a-chip have permeated every segment of thedigital market, from servers and supercomputers, to desktops and smartphone/mobile devices. Innovative applications continue to emerge thatintegrate user interactivity via mobile devices with the power ofhigh-end computation and large-scale storage. These applications tendto demonstrate dynamic and diverse execution behaviors, requiringclose tracking in order to manage resource consumption. Inherentapplication behavior variations and resource requirements, coupledwith complex interactions due to contention for shared resources,present new challenges for efficient, dependable, and sustainableadvances in computing and information technologies.This project will develop foundational system mechanisms to manageindividual cores, memory, and power/energy consumption onmulticore-based systems, and explore resource allocation policies thatuse these mechanisms. A new online multicore power model will allowpower and energy tracking at a fine-grain level, combining availablehardware-based power measurements with event-based modeling andattribution of power for shared resources. Accuracy will be improvedvia coarse-grain measurement-triggered online model recalibration. Anew operating system mechanism, power containers, will be developed toallow application-defined boundaries for resource tracking, breakingfrom the traditional thread and process boundary resource tracking instate-of-the-art operating systems. Further optimizations will beexplored to lower the overhead of resource tracking, allowing onlinepower accounting to scale to large core and resource counts. Finally,new techniques will be developed to enable use of the power containersto control and isolate power usage in an application-defined manner.This project will enhance the power protection, energy efficiency, anddependability of multicore-based computer systems. The techniquesdeveloped can be broadly applied to emerging dynamic applications indata centers, desktops, and mobile devices. In particular, powercontainers will identify and mitigate (either malicious orunintentional) power anomalies (execution that results in unusuallyhigh power consumption), in both high-end servers and multi-core andmulti-accelerator smart phones. The success of this project willcontribute to the long-term sustainability of the world's fastevolving digital economy and society.