Concurrent Maintainability Considerations for Liquid-Cooled Deployments

Author: Jabari George, J.M. Gross Engineering; Ryan Orr, Uptime Institute (Topology Services) Collection: 2023 Global Summit - Cooling Environments

Summary

This presentation, part of the OCP Advanced Cooling Facilities (ACF) work stream, addresses the practical challenge of achieving concurrent maintainability (CM) in liquid-cooled data center deployments. Jabari George (J.M. Gross Engineering) and Ryan Orr (Uptime Institute / Topology Services) collaborated with OCP members across the value chain to produce a white paper framework addressing how to design and operate liquid-cooled data centers that meet the concurrent maintainability standard.

Ryan Orr begins by establishing the precise definition of concurrent maintainability: "the capability to isolate, maintain, remove, and repair any capacity component, distribution pathway, or other element for planned events without impacting the power, cooling, or telecommunications to the critical IT equipment." This requires permanently installed redundancy — temporary or operational workarounds do not satisfy the definition. Concurrent maintainability is typically considered the minimum performance benchmark for continuous application availability (equivalent to Uptime Institute Tier III).

The presentation draws a crucial distinction between achieving CM at the individual rack level (the hardest scenario, which the white paper focuses on) versus at the system level (e.g., two racks in separate availability zones, where only one needs to remain operational). As liquid cooling becomes more mainstream — moving beyond HPC applications with high downtime tolerance into mission-critical environments — the industry needs clear frameworks for how to implement CM with DLC. The white paper examines three liquid cooling topologies (immersion, cold plate, door heat exchanger) against CM considerations. Each topology has unique challenges: immersion involves a fluid as potential single point of failure, level management complexity, and constant air/fluid interface; cold plate has non-redundant cold plates and requires filtration attention; door HX is most analogous to air cooling in terms of maintenance approaches but has single-unit failure exposure.

A key practical issue raised is the IT/facilities boundary — when liquid cooling systems extend closer to the chip, the line between IT-owned and facilities-owned equipment blurs, creating operational confusion about who is responsible for which maintenance activities. The white paper includes a spreadsheet tool to manage this interface.