服务器
沉浸式(数学)
计算流体力学
冷却液
数据中心
模拟
热的
计算机科学
机械工程
工程类
操作系统
热力学
物理
数学
纯数学
航空航天工程
作者
Satyam Saini,Jessica Gullbrand,Suchismita Sarangi,Eric McAfee,Drew Damm
出处
期刊:I-THERM
[Institute of Electrical and Electronics Engineers]
日期:2023-05-30
卷期号:: 1-7
被引量:5
标识
DOI:10.1109/itherm55368.2023.10177563
摘要
Thermal management of data center platforms is becoming more challenging due to rising performance demands accompanied by increasing platform power densities. To address this issue, advanced liquid-based cooling technologies are being adopted in modern data center deployments. One of these liquid cooling technologies is single-phase immersion. Single-phase immersion cooling promises to be a viable and cost-effective approach to manage the cooling requirements of high-performance servers and enables cooling efficiency and sustainability at the same time. While it is known that the servers will exhibit superior thermal performance in immersion compared to air, a lot of unknowns also arise when a server is submerged in a tank. Some of these fundamental unknowns are determining homogeneity in the delivery of the coolant flow from the tank to each of the immersed servers and what are the most efficient tank fluid delivery designs to the servers for optimal thermal cooling capability. The present investigation aims at developing a numerical modeling technique using CFD to analyze server performance in single-phase immersion cooling for a simplified server model in an immersion tank. An in-depth assessment of the influence of tank fluid delivery mechanism on the thermal performance of the server is analyzed. The issue of flow rate maldistribution among individual servers in the tank and quantification of flow bypass around the servers is assessed by varying server inlet flow boundary conditions. A method to quantify the influence of the fluid delivery mechanism at the server level is presented. A discussion on potential flow delivery challenges to the high-power components on the server in immersion and best server flow delivery design practices derived from the simulation results is also described.
科研通智能强力驱动
Strongly Powered by AbleSci AI