SHF: Small: Taming the Combinatorial Explosion of Power Management for Future Manycore Systems

我们将很快进入多核时代，在这个时代，系统将包括数百个核心、大型多级缓存结构、复杂的片上网络、许多内存控制器和大量的主内存。这些系统还将包含丰富的电源管理机制，并将努力在各种功率、能源和热量限制下实现高性能。遗憾的是，系统组件的电源管理不协调可能会导致振荡行为、更高的功率/能量/温度和/或性能下降。鉴于其大量的硬件组件和广泛的可用机制，许多核心系统将不得不协调其不同电源管理器的行动。此外，为了确定(协调的)路线，这些系统将必须全面而严格地对各种性能和功率/能量/温度权衡进行推理。该项目将开发全球功率协调(GPC)来管理多个核心系统中可能的功率管理配置的组合爆炸。GPC将使用一个引擎来实现，该引擎分析整个系统可能的功率状态配置空间。然后，它将决定哪些配置是最合适的，并使用每个组件的电源管理器来执行适当的设置。为了优化搜索良好的配置，GPC将考虑贪婪的搜索启发式算法，该算法依赖于估计电源状态变化的好处的新技术，具有类似电源管理挂钩的分组资源，并利用先前观察到的行为和分级组织。该项目将在许多方面广泛影响社会。首先，解决许多核心系统的电力、能源和温度问题可以极大地影响运行互联网服务的数据中心和推动我们科学发展的高性能系统。其次，解决这些问题将解决计算机行业的关键技术障碍之一。第三，该项目将教育研究生、本科生和高中生，同时扩大未被充分代表的群体对计算机科学的参与。

We will soon enter the manycore era, in which systems will includehundreds of cores, large multi-level cache structures, sophisticatednetworks-on-chip, many memory controllers, and huge amounts of mainmemory. These systems will also embody a rich array of powermanagement mechanisms and will strive to achieve high performanceunder various power, energy, and thermal constraints. Unfortunately,the uncoordinated power management of a system's components canproduce oscillating behavior, higher power/energy/temperature, and/orexcessive performance degradation. Given their large number ofhardware components and the wide spectrum of available mechanisms,manycore systems will have to coordinate the actions of their variouspower managers. Moreover, to decide on a (coordinated) course ofaction, these systems will have to comprehensively and rigorouslyreason about various performance and power/energy/temperaturetradeoffs.This project will develop global power coordination (GPC) to manage thecombinatorial explosion of possible power management configurations inmanycore systems. GPC will be realized using an engine that analyzesthe space of possible power state configurations for the entiresystem. It will then decide which configurations are most appropriate,and use per-component power managers to actuate the proper settings.To optimize the search for good configurations, GPC will considergreedy search heuristics that prune the space by relying on noveltechniques that estimate the benefit of power state changes, groupresources with similar power management hooks, and leverage priorobserved behavior and a hierarchical organization.This project will broadly impact society in many ways. First,addressing the power, energy, and temperature problems of manycoresystems can greatly impact the datacenters that run our Internetservices and the high-performance systems that advance our science.Second, tackling these problems will address one of the keytechnological barriers in the computing industry. Third, the projectwill educate graduate, undergraduate, and high-school students, whilebroadening the participation of underrepresented groups in computerscience.