Start here
What evidence makes a ride-through claim reviewable?
Ask for the failed component or loop, supported load, initial conditions, continuing services, start and end definitions, and the modeled or tested basis. Then ask a qualified reviewer whether that evidence applies to the installed configuration. A supplied interval is evidence for its stated scenario; it is not a general safe-runtime allowance or permission to interrupt cooling.
Write the interruption as a complete sentence
Open Compute Project's liquid-distribution guidance describes thermal ride-through as an interval of continued IT operation following degraded cooling performance or lost facility heat transfer. It separately notes overheating risk from loss of flow in liquid loops and places ride-through calculations within professional engineering consideration. Those distinctions make the failure description essential.
Use the site's own loop names and drawing references. Expand facility water system (FWS) and technology cooling system (TCS) when first used in the register, then identify the actual circuit. A sentence such as “the facility-side heat-transfer service is unavailable while the specified technology-loop circulation continues” is more reviewable than “cooling fails.” It still needs the supporting document and conditions.
List the services the claim assumes remain available. Ask its author to identify any pump, power, control, heat-rejection or IT behavior included in the analysis. This is a request to explain the supplied claim, not a proposed operating configuration.
References: [1]
Read the conditions behind the interval
Locate the exact statement in the report, including its units and qualifiers. Identify whether the interval begins at an assumed failure, a detected event or another defined point. Record the endpoint in the author's own technical terms and have its meaning explained for the decision-maker. “Time until alarm” and “time until an equipment limit” are different statements even when both use seconds.
Preserve the load basis and starting state. A report for one rack configuration cannot be silently relabeled as evidence for a larger deployment. A graph without the model version, configuration and assumptions may be useful background, but it leaves the applicability question unanswered.
| Evidence field | What to record | Unresolved if missing |
|---|---|---|
| Failed boundary | Named loop/component, event and drawing revision | Which interruption was evaluated? |
| Load and starting state | IT configuration, load basis and initial conditions | Does the present installation match? |
| Continuing services | Circulation, power, controls and other stated assumptions | What does the claim depend on? |
| Time definition | Start event, endpoint, interval and units | What does the quoted duration actually mean? |
| Evidence and review | Model/test record, limitations and named qualified reviewer | Who has established applicability? |
Worked example: 300 seconds and 45 seconds do not rank resilience
Consider two fictional reports for a procurement discussion. Report A models an interruption of facility heat transfer while technology-loop circulation remains available. It reports 300 seconds to its specified modeled endpoint at a stated 400 kW IT load. Report B records a laboratory observation after technology-loop circulation stops, with 45 seconds to a different stated endpoint at 250 kW. These invented values are not operating allowances or Airplex test results.
A buyer cannot conclude that A provides 255 seconds more protection. The failed loop, load, evidence type and endpoints differ. Subtracting the numbers is valid arithmetic but gives no defensible comparison of the installations. Even a common IT load would leave the failure and endpoint differences unresolved.
The useful review outcome is two separate entries. Ask each author to explain applicability to the owner's actual scenario and identify the additional evidence needed for a common comparison. If the concern is loss of technology-loop circulation, Report A's longer figure does not answer it. Retain it for the narrower scenario it describes.
| Field | Claim A | Claim B | Review consequence |
|---|---|---|---|
| Interruption | Facility heat transfer lost; TCS circulation continues | TCS circulation stops | Different failed boundaries |
| Load | 400 kW stated IT load | 250 kW stated IT load | Different evaluated duties |
| Time and basis | 300 s; model | 45 s; laboratory record | Different evidence and configurations |
| Endpoint | Model's defined endpoint A | Test record's defined endpoint B | Obtain definitions; do not rank the numbers |
Separate evidence quality from applicability
A modeled result should identify the model, assumptions, inputs and any validation the author relies on. A test record should identify the tested configuration, observed conditions, time record and limitations. Asking for these items does not prescribe a test or favor one evidence type automatically. It lets the qualified reviewer explain what the record establishes.
A detailed laboratory report can still concern different hardware. A carefully documented model can still use assumptions that the installed site does not meet. Record both judgments: what supports the reported result, and what supports using it for this particular decision. Do not compress them into one checkbox labeled “tested.”
For a multitenant facility, ask who can substantiate the building-side assumptions and who owns the IT-side assumptions. Keep confidential system details in the agreed project channel. A supplier presentation and a landlord statement may describe different system boundaries; the reviewer needs the connection between them.
Close the review with an explicit disposition
Useful outcomes include “applicable to the stated scenario,” “applicable only with the listed conditions,” or “applicability unresolved.” Have the qualified reviewer supply the wording and reference. Keep the interval, configuration and exclusions together so a later reader cannot copy only the number into a response plan.
A change to racks, load, loop arrangement, controls or the stated supporting services should prompt the responsible owner to revisit the evidence. Preserve the earlier record as history. Do not reuse its date or approval for a changed configuration without review.
Use the redundancy guide to examine available paths and the tabletop guide to examine people's decisions. Neither establishes ride-through time. Any proposed physical demonstration belongs in a separately authorized technical scope; this page provides no interruption sequence, safe runtime calculation or recovery procedure.
Put this guide to work
Thermal ride-through claim register
Compare two supplied claims without changing equipment. Record exact report references and use the last column for the qualified reviewer's applicability decision. Blank cells are for your evidence; example durations are deliberately excluded.
Open each item to add your notes, then download the CSV to keep a copy. This tool does not save or submit your entries; do not rely on your browser restoring them.
Failed loop/component / event
IT configuration / evaluated load
Initial temperatures / flow / state
Continuing services and assumptions
Start / endpoint / claimed interval
Modeled/tested basis / limitations
Site applicability / review disposition
Continue with a related decision
Questions & answers
Does N+1 prove a particular ride-through time?
No. A capacity or path arrangement does not supply an interval for a defined interruption. Request the time claim's scenario, assumptions and supporting analysis. Keep the redundancy description and ride-through evidence linked but distinct.
Can we use a tank's water volume to calculate safe runtime?
This page does not provide that calculation. The responsible engineers must evaluate the actual system, including which heat and flow paths remain available in the named event. A volume figure alone cannot establish permission to keep IT operating after cooling is interrupted.
What if a report quotes minutes without a defined endpoint?
Keep the claim unresolved and ask its author what starts and ends the interval. Preserve the original statement rather than supplying your own interpretation. The site's qualified reviewer then decides whether the clarified evidence answers the intended question.
Sources & technical references
- OCP: Data Center Liquid Distribution Guidance & Reference Designs
Revision 1.0, January 8, 2024. Thermal Ride-Through discussion on page 10 read November 16, 2026. Supports the distinction between degraded heat transfer and lost loop flow, and professional engineering review; supplies no site runtime here.
NYC & Northern New Jersey
Identify the claim that needs review
Airplex Mechanical's Critical Cooling Division supports liquid-cooling and CDU projects across NYC and Northern New Jersey. Bring the equipment details, supplied time claim and decision needed to discuss the mechanical work. Specialized modeling, test work and engineering acceptance must be explicitly assigned in the project scope.
Discuss the cooling scopeCall (347) 627-7146