Scholarlink

<p>Air-assisted liquid cooling (AALC) provides a pragmatic solution for datacenters and colocation ecosystems which houses conventionally air-cooled server racks. This approach minimizes retrofitting requirements, and the easy integration associated with this design allows for large-scale deployment alongside standard air-cooled systems. The computation demand for AI hardware deployed comes with high amounts of power consumption (30–100&#xa0;kW per rack) that operates with high variability. This inevitably produces an incredible amount of heat, which needs to be efficiently removed from the compute processors to ensure reliability, longevity, and minimal downtime. AALC systems effectively manage high thermal design power (TDP) processors, overcoming the limitations of air cooling, which struggles with heat dissipation in high-performance environments. Liquid cooling comes with its own set of issues and failure mechanisms to understand. This paper highlights the issues encountered, failure analysis performed, root cause, and corrective actions that were enabled on AALC through long-term reliability (LTR) testing.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Failure Analysis of Liquid Cooling Setup for Hyperscale Datacenter Infrastructure

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Pradip Sairam Pichumani,
  • Ting Yan,
  • Eoghan Dillon

摘要

Air-assisted liquid cooling (AALC) provides a pragmatic solution for datacenters and colocation ecosystems which houses conventionally air-cooled server racks. This approach minimizes retrofitting requirements, and the easy integration associated with this design allows for large-scale deployment alongside standard air-cooled systems. The computation demand for AI hardware deployed comes with high amounts of power consumption (30–100 kW per rack) that operates with high variability. This inevitably produces an incredible amount of heat, which needs to be efficiently removed from the compute processors to ensure reliability, longevity, and minimal downtime. AALC systems effectively manage high thermal design power (TDP) processors, overcoming the limitations of air cooling, which struggles with heat dissipation in high-performance environments. Liquid cooling comes with its own set of issues and failure mechanisms to understand. This paper highlights the issues encountered, failure analysis performed, root cause, and corrective actions that were enabled on AALC through long-term reliability (LTR) testing.