DESC: Type I: Towards Greener AI Computing: Designing and Managing Sustainable Heterogeneous Edge Data Centers

人工智能（AI）最近成为计算研究的一个突出领域，具有对社会产生积极影响的巨大潜力，预计在未来几年将大幅增长。人工智能市场预计将在未来5年内扩大10倍，成为一个价值数千亿美元的行业。然而，人工智能技术的快速发展已经造成了相当大的环境影响，并引发了人们对其相关运营和隐含碳足迹的担忧。前者与设备的持续运行和维护有关，而后者与设备的整个生命周期有关。该项目旨在通过减轻人工智能对环境的影响和提高数据处理效率，大幅提高现代边缘数据中心的可持续性。它将超越人工智能，朝着可持续边缘数据中心的发展迈进。它还将促进整个计算机行业采用可持续发展的做法。因此，它将有助于更广泛的范式转变，在计算领域采用更环保的实践，并将加强努力，推动数字基础设施向更绿色、更可持续的未来过渡。该项目的核心目标是通过广泛的多层优化来提高用于人工智能计算的边缘数据中心的可持续性。与传统的集中式数据中心不同，边缘数据中心的设计更接近最终用户，并且可以利用可再生能源进行运营。这最大限度地减少了对大量数据传输的需求，并支持更加本地化的计算基础设施。该项目超越了传统的优化方法，以垂直解决方案为目标，因为多个参数会影响操作和隐含的碳足迹。更具体地说，可以通过利用可再生能源、提高电力使用效率和战略性地在各种计算组件之间分配工作负载来减少操作碳足迹。另一方面，可以通过缩小节能硬件加速器的规模、实施低足迹电路设计以及提高系统可靠性来延长其使用寿命来减少所包含的碳足迹。为了涵盖所有这些方面，该项目将遵循碳优先的垂直方法，最初将为边缘数据中心开发强大的电路级和架构级硬件加速器，然后将提高系统的利用率，并最终在运行时使用可再生能源有效地管理边缘数据中心。总体而言，开发可靠的硬件加速器和有效利用计算资源，将通过减少碳足迹和提高整体可持续性，带来更节能的数据中心。此外，在边缘数据中心的管理中整合可再生能源将促进清洁能源的使用，减少对不可再生能源的依赖。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Artificial Intelligence (AI) has recently emerged as a prominent field of research in computing, with enormous potential to positively influence society and it is projected to grow considerably in the coming years. The AI market is anticipated to expand tenfold into a sector worth hundreds of billions of dollars within the next five years. However, the rapid expansion of AI technologies has resulted in considerable environmental impact and raised concerns about their associated operational and embodied carbon footprints. The former is associated with their ongoing operation and maintenance while the latter with the entire life cycle of the devices. This project aims to substantially improve the sustainability of modern edge data centers by mitigating the environmental impact of AI and boosting the efficiency of data processing. It will advance beyond AI, towards the development of sustainable edge data centers. It will also promote the adoption of sustainable practices in the computing industry as a whole. As a result, it will contribute to a broader paradigm shift towards embracing more eco-friendly practices in the computing domain and it will strengthen efforts in driving the digital infrastructure’s transition towards a greener and more sustainable future.The central aim of this project is to improve the sustainability of edge data centers used for AI computing through extensive, multi-layered optimizations. Unlike conventional centralized data centers, edge data centers are designed to be closer to end-users and can harness renewable energy for their operations. This minimizes the need for extensive data transmission and enables a more localized computing infrastructure. This project moves beyond traditional optimization methods and targets vertical solutions as multiple parameters affect operational and embodied carbon footprints. More specifically, the operational carbon footprint can be lessened by harnessing renewable energy sources, enhancing power usage efficiency, and strategically distributing workload across various computing components. On the other hand, the embodied carbon footprint can be decreased by downscaling energy-efficient hardware accelerators, implementing low-footprint circuit designs, and enhancing system reliability to extend its lifespan. To cover all these aspects, this project will follow a carbon-first vertical approach that will initially develop robust circuit-level and architectural-level hardware accelerators for edge data centers, then will improve the system’s utilization, and will eventually efficiently manage edge data centers at the run-time level with renewable energy sources. Overall, the development of reliable hardware accelerators and efficient utilization of computing resources will lead to more energy-efficient data centers, by reducing their carbon footprint and improving their overall sustainability. Furthermore, integrating renewable energy sources in the management of edge data centers will promote the use of clean energy and reduce reliance on non-renewable sources.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.