White paperPDF2022

Leak Detection and Intervention, Revision 1.0 (OCP Advanced Cooling Solutions white paper)

Open Compute Project Foundation · Michael Berktold, Jessica Gullbrand et al.

An educational OCP white paper that compares leak risk, detection methods and manual and automatic intervention options, with pros and cons, for the Technology Cooling System (TCS) of liquid-cooled IT.

CC BY-SA 4.0 — share with attribution under the same licence

On this page

Overview

The paper is scoped to the TCS, meaning cold plates, integrated pumps, manifolds, CDUs, heat exchangers, hoses and couplings, and covers water with additives, glycol-based, dielectric and refrigerant coolants. Air struggles at very high power densities while liquid cooling brings leak concerns near sensitive electronics. The document is educational guidance, not a normative specification; it recommends no specific solution and does not go deep on system management. It is meant to be used with OCP's cold plate requirements and rear-door heat exchanger documents.

The design chapter starts with designing for no leaks: a risk analysis (probability and impact) and total-cost-of-ownership evaluation at project start. For serviceable connections, BSPP with an O-ring or washer is preferred over BSPT with tape or putty; non-serviced joints can be brazed or welded, and containment pans considered in high-risk areas. National-lab lessons are given: a cheap O-ring caused slow leaks too small to trip the facility alarm, leading to repositioned detection cable, a change from two-wire to four-wire rope, and seal fatigue and side-load problems in quick disconnects.

The commissioning chapter describes gas pressure testing. Helium vacuum-chamber testing suits manufacturing because helium is the smallest molecule; dry nitrogen is used in the field, and a standard air compressor must not be used. In the gas dwell test the system is pressurised with dry nitrogen slightly below the relief valve setting (for example 35 psi) for 8 hours with no pressure drop. When filling the facility water system, 14.5 psi (1 bar) static pressure is held for one hour with no loss. For refrigerant systems, citing ANSI/ASHRAE 34-2019, at least 1,000 ft3 (28.3 m3) of space is needed per 13 lb of refrigerant.

The detection chapter separates indirect and direct methods. Indirect methods are non-contact: differential pressure across the TCS pumps over time, reservoir level and optical sensors; differential pressure measurement is suggested as a requirement and is typically available in off-the-shelf CDUs. Direct detection needs the liquid to touch the sensor and to be conductive; the common solution is a two-wire (leak present) or four-wire (location) rope, plus point sensors and catch pans. In slow leaks water evaporates and leaves glycol residue that a rope can miss. Alarms reach the DCIM or BMS via Modbus RTU/IP, SNMP or BACnet; Redfish liquid-cooling schemas had not yet been accepted by DMTF.

Severity is classified as minor (no visible spill, indirect detection), major (visible within IT equipment, single-zone direct detection) and critical (pooling outside the rack, multi-zone detection). Intervention is graded from notification only, to electrical de-energisation, to de-energisation plus fluid shut-off. Manual tools are shut-off valves, dry-break quick disconnects (UQD2 drips less than 0.063 cc at 200 psi) and interlocking valves; automatic tools are solenoid valves and CDU shutdown. Before a CDU shutdown the affected IT should get a safe shutdown, and cold-plate processors must have thermal throttling enabled. Redundant-sensor logic (one sensor alarms, two trigger automatic action) reduces nuisance alarms.

Key points

  1. For serviceable connections BSPP with an O-ring or washer is preferred over BSPT with tape or putty; non-serviced joints can be brazed or welded and containment pans used in high-risk areas.
  2. Gas dwell test: pressurise with dry nitrogen slightly below relief valve setting (for example 35 psi) for 8 hours with no pressure drop.
  3. When filling the facility water system, hold 14.5 psi (1 bar) static pressure for one hour with no pressure loss.
  4. For refrigerant systems, per ASHRAE 34-2019 as cited, at least 1,000 ft3 (28.3 m3) of space volume is needed per 13 lb of refrigerant.
  5. Leak severity is minor (no visible spill), major (visible within IT equipment) or critical (pooling outside the rack, multi-zone detection).
  6. Differential pressure measurement across TCS pumps is suggested as a requirement and is typically available in off-the-shelf CDUs.
  7. A four-wire leak rope gives leak location while a two-wire rope only indicates presence; glycol residue from slow leaks can be missed by a rope.
  8. A UQD2 dry-break quick disconnect is specified to drip less than 0.063 cc when disconnected at 200 psi.
  9. Issue a safe shutdown to IT before the CDU shuts down, and enable thermal throttling on cold-plate processors.
  10. Intervention levels run from notification only, to electrical de-energisation, to de-energisation plus fluid shut-off; cost rises as potential damage falls.

Using it in design

  • Run a probability-and-impact leak risk analysis at project start, specify BSPP/O-ring joints for serviceable connections and check the O-ring compound against the coolant chemistry using the full wetted-materials list.
  • Add an 8-hour dry-nitrogen dwell test (below relief valve setting) and a 1-hour 14.5 psi hydrostatic test for the facility water system to the commissioning procedure, and forbid air compressors as the test gas source.
  • Layer detection: differential pressure monitoring across TCS pumps, rope or point sensors under CDUs and racks, and four-wire location-capable rope under raised floors with rope positions mapped to facility coordinates.
  • Agree alarm thresholds at commissioning, use redundancy logic (either sensor alarms, both trigger automatic action) to cut nuisance alarms, and insulate pipework against condensation.
  • Set the intervention strategy by the customer's redundancy: check solenoid valve power, fail-state behaviour and swing time (water hammer), and implement in the BMS a safe-shutdown signal to IT ahead of any CDU shutdown.
  • Where a redundant CDU exists, specify that the BMS sends the shutdown signal to both CDUs.

Monthly newsletter · free

Keep your reading list current

Once a month: new library sources, changes to Turkish regulations and upcoming data centre events.

SoSo CAD

Check cooling capacity alongside your design

Review zone heat loads, available capacity and the capacity balance after the largest unit is lost in SoSo CAD.

Request a demo
Zone heat loads and capacity balance in the SoSo CAD cooling workspace

Who should read it

Mechanical and BMS engineers designing direct liquid-cooled racks and rows, and operators writing leak detection and response requirements.

Document contents

  1. License
  2. Executive Summary
  3. Introduction
  4. Design Considerations
  5. Definitions
  6. Examples
  7. Risk Analysis of an example
  8. Lessons learned from early liquid cooling applications at US National Labs
  9. Reliability Data
  10. Gaseous Pressure Testing
  11. Leak Detection and Mitigation
  12. Indirect
  13. Direct
  14. Installations
  15. Control Systems
  16. Reporting
  17. Nuisance Alarms
  18. Leak Intervention
  19. Manual (shutoff valves, dry break quick disconnects, interlocking valves)
  20. Automatic (electrical de-energization, solenoid valves, CDU shut downs)
  21. Conclusion

Read the document

ocp-leak-detection-whitepaper.pdf

PDF · 13.5 MB · English

CC BY-SA 4.0 — share with attribution under the same licence

The document is provided unaltered because its licence permits redistribution.

Publication details

Authors
Michael Berktold; Jessica Gullbrand; Fred Rebarber; Nigel Gore; Matthew Archibald; Michael Thompson; Cam Turner
Publisher
Open Compute Project Foundation
Year
2022
Document language
English
Source type
Industry body
Access
Open licence — downloadable
Licence
CC BY-SA 4.0 — share with attribution under the same licence
Licence evidence
www.opencompute.org
Original licence note

License section of the document states Creative Commons Attribution-ShareAlike 4.0. Attribution required; derivatives/adaptations must stay CC-BY-SA.

Citation

Michael Berktold; Jessica Gullbrand; Fred Rebarber; Nigel Gore; Matthew Archibald; Michael Thompson; Cam Turner (2022). Leak Detection and Intervention, Revision 1.0 (OCP Advanced Cooling Solutions white paper). Open Compute Project Foundation. https://www.opencompute.org/documents/acs-cold-plate-leak-detection-and-intervention-white-paper-pdf-1

All content and rights belong to the respective authors and publishers. SoSo CAD provides only the citation, its own summary and a link to the source, and is not responsible for the sources' content.