Collaborative Research: NeTS-NBD: Increasing the Energy Efficiency of the Internet with a Focus on Edge Devices

该提案通过关注主要且经常被忽视的能源消费者——边缘设备，解决了提高互联网能源效率这一日益重要的需求。劳伦斯伯克利国家实验室（LBNL）的研究表明，仅在美国，互联网就消耗了大约74太瓦时/年的电力（每年约60亿美元），其中24太瓦时/年或32%的电力可以在台式电脑上充分利用电源管理，目前是互联网上最常见的边缘设备。不幸的是，由于现有协议和体系结构的限制，联网桌面计算机通常在频繁且经常长时间空闲期间保持通电状态。作为网络设备，由于它们需要在没有警告的情况下响应各种类型的网络交易，因此无法以节能的方式运行。NSF目前正在资助研究降低一级局域网交换机、服务器集群和超级计算机能耗的项目，但严重缺乏降低互联网能耗最大部分——边缘设备能耗的研究。解决这一挑战的主要研究人员（pi）方法是研究和利用一套基于流量和系统约束的协议和子系统基础设施的新研究概念，以及有效控制它们的算法，以便桌面和其他边缘设备可以在相对空闲期间进入睡眠状态，同时通过集成到系统中的低功耗硬件代理来维持网络连接。此外，PIs的方法还承诺在活动期间减少与网络相关的资源消耗，从而提供额外的能源效率提高，在活动期间，性能的优雅下降是可以接受的，实际上是以速度换取能源。除台式计算机外，该方法还将为各种新兴的有线和无线边缘设备（如电视机顶盒、网络设备、远程摄像机等）提供类似的解决方案。该方法中的新概念包括：(1)在台式计算机和/或一级局域网交换机的网络接口中使用协议代理，以减少边缘设备的系统资源的少量使用，从而允许系统电源管理得到充分利用；(2)“智能”唤醒方法，允许仅在现有应用程序和协议需要时透明地唤醒电源管理设备，并为未来的网络应用程序提供新的电源管理通知语义；(3)以太网中的自适应链路速率，通过以不同的数据速率动态操作链路（例如，仅在需要时为1或10 Gb/s，否则为10或100 Mb/s）来权衡性能（或QoS）以提高能源效率；(4)用于固定或可重构的阶段硬件功能级别的架构，通过自适应的空间或时间分配硬件资源来实现节能运行，并具有功率、性能和功能缩放。所建议的实现这些节能的方法是对新软件和硬件进行测量、建模和原型设计。pi将构建和评估基于自适应功能级别的边缘设备高级以太网接口原型，包括代理、智能唤醒、自适应链路速率和重新配置。他们将为跨网络的电源管理通知使用新语义的软件应用程序设计原型。他们将把研究成果传播给标准组织、政府机构和/或行业。pi将设计和评估适合IEEE 802.3考虑的自适应链路速率的快速自协商机制。他们还希望在影响未来EPA能源之星对台式电脑和其他边缘设备（如潜在的电视机顶盒）的规范方面发挥关键作用。该项目将在南佛罗里达大学和佛罗里达大学的合作伙伴关系下进行。将支持四名研究生，其中一名可能来自代表性不足的人口群体。智力价值：所提出的工作在智力上是有价值的，因为它是全球互联网在其主导因素（网络连接的边缘设备）方面实现节能的第一次重大努力。这项研究将定义网络边缘设备能源效率的重要性，作为经济和环境问题，并导致电源管理和网络应用，协议和架构的集成设计。更广泛的影响：这项工作的更广泛的影响是三重的。首先是在这个项目完成后的几年内对社会的影响，使美国和国外的能源成本显著降低，并通过降低运营成本支持互联网向发展中国家的扩展。如果现有的电源管理功能可以基于本提案中的新想法在网络边缘设备中启用，美国每年将节省数亿至数十亿美元。ppi与LBNL、EPA和工业界密切合作，传播研究成果，实现预期的节能目标。第二个更广泛的影响是对教育过程的影响，无论是直接参与这个项目的研究生，还是将从这项研究将创建和共享的材料中受益的许多大学生。最后，通过NSF REU和RET计划，将向K-12和代表性不足的人口进行推广。

This proposal addresses the increasingly critical need to improve the energy efficiency of the Internet by focusing on the primary and often neglected energy consumer, edge devices. Studies by Lawrence Berkeley National Laboratory (LBNL) show that about 74 TWh/yr of electricity (which is approximately $6 billion per year) is consumed by the Internet in the USA alone, of which 24 TWh/yr or 32% could be saved with full use of power management on desktop computers, currently the most common of edge devices on the Internet. Unfortunately, due to limits of existing protocols and architectures, networked desktop computers typically remain powered-up during frequent and often lengthy periods of idleness. As network devices, they are prevented from operating in an energy-efficient manner due to their need to respond to network transactions of various types without warning. The NSF is currently funding projects to investigate energy-use reduction in first-level LAN switches, server clusters, and supercomputers, but severely lacking is research to reduce energy consumption of the largest portion of Internet energy consumers, the edge devices.The principal investigators'(PIs) approach to addressing this challenge is to investigate and exploit a synergistic set of novel research concepts for protocol and subsystem infrastructure, and algorithms for effectively controlling them based on traffic and system constraints, so that desktop and other edge devices can be put to sleep during periods of relative idleness while network connectivity is maintained by a low-power hardware proxy integrated into the system. Moreover, the PIs' approach also promises to provide additional increases in energy efficiency by reducing consumption of network-related resources during active periods where graceful degradation of performance is acceptable, in effect trading off speed for energy. In addition to desktop computers, this approach will lead to similar solutions for a wide variety of emerging wired and wireless edge devices such as television set-top boxes, network appliances, remote cameras, etc.The novel concepts in the proposed approach include: (1) protocol proxying in the network interface of a desktop computer and/or within a first-level LAN switch to reduce minor use of system resources of the edge device and thus allow system power management to be fully exploited; (2) "smart" wake-up methods to allow power-managed devices to be awoken transparently and only when needed by existing applications and protocols, along with new power-management notification semantics for future network applications; (3) adaptive link rate in Ethernet to trade-off performance (or QoS) for energy efficiency by dynamically operating links at varying data rates (e.g., 1 or 10 Gb/s only when needed, and otherwise at 10 or 100 Mb/s); and (4) architectures for fixed or reconfigurable levels of staged hardware functionality to realize energy-efficient operation via adaptive spatial or temporal assignment of hardware resources to network transactions with power, performance, and functionality scaling.The proposed methodology to achieve these energy savings is measurement, modeling, and prototyping of new software and hardware. The PIs will build and evaluate a prototype of an advanced Ethernet interface for edge devices based on adaptive levels of functionality including proxying, smart wake-up, adaptive link rate, and reconfiguration. They will prototype software applications that use new semantics for power management notification across a network. They will disseminate the research outcomes to standards groups, government agencies, and/or industry. The PIs will design and evaluate a fast autonegotiation mechanism for adaptive link rate suitable for consideration by the IEEE 802.3. They also expect to play a key role in influencing future EPA Energy Star specifications for desktop computers and other edge devices (such as, potentially, television set-top boxes). This project will be conducted in a collaborative partnership between the University of South Florida and the University of Florida. Four graduate students will be supported of which one may be targeted to be from an underrepresented population group.Intellectual merit: The work proposed is intellectually meritorious in that it is the first significant effort to achieve energy savings for the global Internet in terms of its dominant factor, the network-connected edge devices. This research will define the importance of energy efficiency of network edge devices as an economic and environmental issue and result in the integrated design of power management and network applications, protocols, and architectures. Broader impact: The broader impact of this work is threefold. First is the impact upon society within a few years of completion of this project, enabling significant reductions in energy costs here in the USA and abroad, and supporting the expansion of the Internet into the developing world by reducing operating costs. Savings in the hundreds of millions to billions of dollars per year in the USA will be achieved if existing power management capabilities can be enabled in network edge devices based on the new ideas in this proposal. The PIs have existing relationships to allow them to work closely with LBNL, EPA, and industry to disseminate research outcomes and achieve the expected energy savings. The second broader impact is on the educational process, both in terms of the graduate students directly involved in this project as well as the many university students that will benefit from the materials that will be created and shared from this research. Finally, through the NSF REU and RET programs, outreach will be made to K-12 and underrepresented populations.