DESC: Type I: Dynamic Optimization for Leaps in Sustainability-Centric Computing System Design

计算系统的制造、操作和处理对环境的影响越来越大。传统上，计算系统的设计是为了最大限度地提高能源效率和工作性能，而温室气体排放和电子废物等因素则继续呈上升趋势。这些因素依赖于计算系统的潜在动态，而时间敏感反应具有解决这些问题的潜力。该项目利用动态系统优化解决计算系统的可持续性问题，为设计未来的计算系统明确关注关键的可持续性指标。重点放在推进可持续计算系统的科学理解，并影响未来的关键结构，如数据中心。该项目不仅支持计算领域，而且还提供了设计和控制领域的进步。该项目的成果还通过广泛的多媒体资源开发和学生培训支持科学界和公众的教育。本研究确定了三个目标。第一个目标是改进和重新制定可持续计算系统的模型，以便在优化中集成，通过开发一种将数据驱动的代理模型转换为动态分析友好格式的方法来解决。第二个目标是理解和量化可持续性指标的不同表示对优化问题解决可行性的影响。这是通过可持续性指标的分类来实现的，用于自动生成可解决的优化问题。第三个目标是开发和评估动态系统优化的能力，以可持续发展为中心的计算系统，使用一个新的框架。验证通过数值和实验案例研究的结合进行，范围从服务器冷却到数据中心设计。通过动态系统优化研究计算系统的潜在贡献是可持续性指标的巨大、离散跳跃，例如减少排放、废热和电子废物。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。

The manufacturing, operation, and disposal of computing systems are increasingly impacting the environment. Traditionally, computing systems are designed to maximize energy efficiency and job performance, while elements such as greenhouse gas emissions and electronic waste continue to trend upward. These factors are dependent on the underlying dynamics of the computing systems, and time-sensitive reactions have the potential to address these issues. This project tackles the sustainability of computing systems using dynamic system optimization, setting an explicit focus on critical sustainability metrics for designing future computing systems. An emphasis is placed on pushing forward the scientific understanding of sustainable computing systems, and to impact future critical structures such as datacenters. This project serves to support not only the computing field, but also provides advancements in the design and controls fields. The outcomes of the project also support education of scientific communities and the public alike through expansive multimedia resource development and student training.Three aims are defined for this research. The first aim is to improve and reformulate models of sustainable computing systems for integration within optimization, addressed by development of a methodology to convert data-driven surrogate models into dynamic analysis-friendly formats. The second aim is to understand and quantify the impact different representations of sustainability metrics have on optimization problem solution feasibility. This is pursued through categorization of sustainability metrics for automatic generation of solvable optimization problems. The third aim is to develop and evaluate the capabilities of dynamic system optimization for sustainability-centric computing systems, using a novel framework. Validation is performed through a combination of numerical and experimental case studies, ranging from server cooling to datacenter design. The potential contributions of studying computing systems through dynamic system optimization are large, discrete jumps in sustainability metrics, such as reductions in emissions, waste heat, and electronic waste.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.