A Novel Scalable Modular Immersion Cooling System Architecture for Sustainable Data Center
Carrie Chen, Jiahong Wu, Jun Zhang, Ying-Shan Lo, Allen Liang, Checa Hung, Nishi Ahuja, Qing Qiao
Abstract
Immersion cooling data centers of both hyperscale and the edge have been emerging rapidly, especially recent massive growth of edge data center. System architecture of immersion cooling data center varies in terms of locality, footprint, design cost, PUE target and so on. Meanwhile, computing at the era of sustainability re-focuses on the benchmark of better performance/power via circular economy-oriented system architecture for ultimate goal of net zero carbon emission. Building and maintaining datacenter sustainable and green with advanced liquid cooling is one of top priorities in the industry. The core value of modular system architecture is to lower data center embodied carbon footprint through the methodology of reusing common building block on demand for system scalability and refreshing specific ingredients or building blocks per their own corresponding lifecycle for system longevity. Immersion cooling, as one of advanced liquid cooling solutions to lower operational carbon footprint for sustainable/green data center, is to have heat extracted closer to its source (such as CPU) with the liquid such as fluorinated or oil to achieve higher heat capacity and lower thermal resistance than air. This paper detail introduces Intel Open IP immersion cooling reference system architecture with scalable modular design of server node, coolant distribution unit, and immersion tank for sustainable data center. Both embodied and operational carbon footprint reductions are evaluated and analyzed with system architecture novelty, design optimization, and experimental verification and validation results. Key learnings from engineering practice are valuable reference for scalable immersion cooling system design of cloud and the edge.