How to Evaluate a Data Center CDU Against IT and Facility Requirements

A CDU’s name does not establish that it can connect a particular server to a particular facility cooling loop.

This hypothetical cold-plate proposal shows which conditions to compare on each side, which labels cannot substitute for specifications, and when to select, exclude, or defer a decision.

Data center deep-dive series

What are the two interfaces a CDU must serve?

Evaluate the IT-side supply and the facility-side connection separately. The U.S. Department of Energy’s *Best Practices Guide for Energy-Efficient Data Center Design* explains that, in many liquid-cooling approaches, a cooling distribution unit (CDU) interfaces with the facility cooling loop while supplying liquid at conditions appropriate for the IT equipment. That description establishes a role, not compatibility with a specific installation.

Consider an explicitly hypothetical proposal: servers use cold plates that transfer heat from chips into flowing liquid, and a candidate CDU is offered to connect them to a facility loop. Cold plates are one form of direct liquid cooling; this example does not describe every liquid-cooling arrangement. Neither the server requirements, the facility conditions, nor the candidate CDU’s operating limits have been supplied.

U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

Which conditions belong in the comparison?

Use the table to request evidence for the same loads and operating conditions on both sides. Every status is **evidence missing** in this hypothetical proposal; it is not a finding that a real product fails.

IT-side liquid

IT equipment or facility evidence needed

Equipment-required fluid type and chemistry

Candidate CDU evidence needed

IT-side supply fluid and supported chemistry

Hypothetical proposal status

Evidence missing; “supports liquid cooling” does not establish compatibility

IT-side operation

IT equipment or facility evidence needed

Equipment-required supply temperature, pressure, flow, and heat load

Candidate CDU evidence needed

IT-side operating range at the required flow and load

Hypothetical proposal status

Evidence missing; no common setpoint can be assumed

Changing loads

IT equipment or facility evidence needed

Initial and future heat loads, including part- and low-load operation

Candidate CDU evidence needed

Operating range under those conditions

Hypothetical proposal status

Evidence missing; a maximum-load figure alone is insufficient

Facility connection

IT equipment or facility evidence needed

Actual loop supply and return temperatures, available flow and pressure, and conditions as load changes

Candidate CDU evidence needed

Required facility-side conditions and heat-transfer capability

Hypothetical proposal status

Evidence missing; an IT-side supply claim does not establish facility compatibility

Boundary between sides

IT equipment or facility evidence needed

Equipment and facility connection and fluid requirements

Candidate CDU evidence needed

Schematic and specifications showing whether the fluids are the same and how the circuits are arranged or separated

Hypothetical proposal status

Evidence missing; the name “CDU” does not establish a heat-exchanger or circuit arrangement

Remaining air cooling

IT equipment or facility evidence needed

Heat captured by liquid and heat left to room air

Candidate CDU evidence needed

Liquid-side heat load the CDU is proposed to handle

Hypothetical proposal status

Evidence missing; CDU capacity does not establish that room air cooling can be removed

The IT-side liquid might be treated water, a glycol-based solution, or a dielectric fluid, depending on the technology; that list does not mean the hypothetical cold plates accept all three. Some liquid-cooled systems also leave part of the IT heat load to air cooling. Request the equipment’s heat split before treating the CDU as a replacement for room cooling.

Can a temperature label fill a missing specification?

No: first identify what the temperature describes. The DOE guide’s example of a medium-temperature chilled-water loop supplied at **55°F or higher** is design guidance, not the hypothetical facility’s measured temperature or a universal CDU requirement. The facility row needs that facility’s own operating conditions.

The guide also describes ASHRAE W17, W27, W32, W40, and W45 as classes whose numbers indicate upper limits, in degrees Celsius, for **server-side liquid supply temperature**. A W label alone does not establish fluid or pressure compatibility, or the facility-side chilled-water conditions.

By contrast, the A1–A4 recommended and allowable temperature envelopes discussed in the guide concern **air-cooled equipment inlet air**. The recommended envelope guides efficient operation while maintaining reliability; the allowable envelope describes boundaries tested for equipment functionality, not reliability. Neither is a CDU liquid-supply specification.

Would the same comparison apply to immersion or prove an efficiency gain?

The questions about interfaces still matter, but this cold-plate table cannot be copied over as an immersion-system specification. The DOE guide distinguishes single-phase immersion, in which a CDU pumps dielectric fluid around electronics, from two-phase immersion, in which vapor transfers heat to an exchanger, condenses, and returns in a passive cycle.

“Immersion cooling” alone therefore does not identify the proposed IT-side circuit or CDU arrangement.

Heat rejection is another separate question. The guide explains that warmer cooling-water conditions can facilitate a water-side economizer and that higher W classes may permit dry-cooler use. Whether either option works depends on the particular facility loop and outdoor conditions. It is not an automatic efficiency benefit of the candidate CDU.

When can the proposal be selected, excluded, or held?

For each table row, record **documented match**, **documented mismatch**, or **evidence missing**. If the equipment requirements, facility-loop data, or CDU operating ranges for either side are absent, hold the compatibility decision. A documented mismatch is grounds to exclude the proposal unless a design change is separately evaluated; missing evidence is not itself proof of a mismatch.

Even matching documents do not complete installation approval. Plan subsequent operating checks around the temperatures, flows, and other values used in the comparison, including the accuracy and calibration status of the instruments. Obtaining the missing documents and comparing them under the same operating conditions turns a product description into a testable interface decision.

Sources

• U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

Related reading

• Immersion Cooling: What a Hypothetical Rack Example Helps You Separate
• Air Cooling vs. Liquid Cooling: Define the System Boundary Before Comparing Efficiency

Keep or Replace Data Center Equipment? Set the Total-Cost Comparison Boundary First

Compare the cost of delivering the same workload and service level over one common, explicitly hypothetical future period—not the price of old equipment against the price of new equipment.

A two-option matrix helps separate equipment spending, operating energy, facility effects and lifecycle work. Without site-specific costs, loads and service-life assumptions, it cannot identify a cheaper option.

Data center deep-dive series

What must both options deliver?

The short answer is to hold the service and comparison period constant, then ask what each option would require to deliver them. The matrix below shows where to put the costs; the energy boundary and missing evidence determine whether those entries can eventually be compared.

Suppose, hypothetically, that an existing fleet and a proposed replacement must handle the same transactions and other work for the next three years. Three years is an example, not a recommended ownership period or an estimate of either fleet’s life.

Specify the work by time of day, including low-load operation, plus the required storage, network capacity and continuity. The fleets need not have identical utilization: the question is whether each can provide the same service. The U.S. Department of Energy (DOE) describes server efficiency in work per watt, but that measure alone does not establish the cost of delivering the service.

Set the start and end of the period and ask whether either option needs repair, an upgrade or another replacement before it ends. State how a purchase made before the start is treated rather than entering its old invoice as a new purchase.

DOE’s Best Practices Guide for Energy-Efficient Data Center Design raises first cost versus life-cycle cost; it does not prescribe an accounting rule for historical purchases, depreciation or residual value. Reliability is a requirement for both options, not a saving to assign to one without evidence about outages and their costs.

U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

IBM — What Is a Data Center? | IBM

IBM — What Is a Green Data Center? | IBM

What belongs in each column?

Use this as a question sheet for the hypothetical on-premises comparison, not as a published cost model or a quotation. Leave monetary entries blank until project information is available. For each row, mark what is common to both options, conditional on a real change, excluded with a stated reason, or unknown.

A common cost can appear consistently in both totals or be set aside when examining only their difference; it is not an option-specific saving.

Equipment and deployment

Retain existing equipment

State the treatment of the earlier purchase; identify necessary upgrades

Replace equipment

Separate acquisition from installation and deployment

Boundary to record

Record future expenditure and timing; leave unquoted amounts unknown

IT electricity

Retain existing equipment

Assess energy for the specified work and load profile

Replace equipment

Assess energy for that same work and profile

Boundary to record

Use the same period and measurement boundary

Storage and networking

Retain existing equipment

Identify required capacity, redundancy, equipment and energy

Replace equipment

Identify anything the refresh would change

Boundary to record

Mark genuinely unchanged shared items as common, not savings

Supporting facility

Retain existing equipment

Check affected cooling, fans, UPS and distribution

Replace equipment

Check load effects and any necessary power or cooling changes

Boundary to record

Include modifications or specialized cooling maintenance only if required

Maintenance and later work

Retain existing equipment

Check repairs, refurbishment, upgrades and component replacement

Replace equipment

Check maintenance and any further intervention within the period

Boundary to record

Record whether and when work is needed; do not assume either option needs none

Service and thermal fit

Retain existing equipment

Check continuity and operating conditions

Replace equipment

Check the same requirements and site cooling capacity

Boundary to record

Establish feasibility before assigning a cost or benefit

End-of-use treatment

Retain existing equipment

Consider continued use, refurbishment or eventual recycling

Replace equipment

Consider the removed equipment and the replacement’s status at period end

Boundary to record

Enter disposal costs or remaining value only with project evidence

A replacement does not automatically require a cooling rebuild, nor does retention mean maintenance-free operation. DOE advises checking initial, future and low-load cooling conditions.

It also distinguishes recommended operating conditions from manufacturers’ allowable functionality boundaries; being within an allowable boundary is not, by itself, a reliability guarantee. IBM’s equipment-lifecycle discussion identifies upgrades, refurbishment, reuse and responsible recycling as possibilities, but provides no prices or dates for this project.

How do you keep electricity costs from being counted twice?

Estimate IT electricity for the same work under each option’s plausible operating-load profile, including part- and low-load periods. Check storage and network energy where those systems change. DOE explains that an IT-load change can also affect cooling and electrical systems, but it does not establish a saving for this particular site.

Record affected facility energy—such as cooling, fans, UPS and distribution losses—without claiming that the shared facility’s entire bill changes with one equipment choice.

Label the meter boundary before adding energy rows. A total-facility electricity figure already includes IT and supporting loads; an IT-only figure does not. Adding cooling or UPS consumption again to the facility total would double-count it.

PUE relates facility energy to IT energy; it is performance context, not a monetary total or an extra multiplier for an already measured facility total. Water and carbon measures likewise need their own boundaries and cost evidence before they become cost entries.

Converting energy into charges requires this facility’s applicable prices and, where relevant, demand charges or contract terms. DOE’s illustrative electricity price is not a tariff for this comparison. Label each energy input measured, supported estimate or unavailable.

DOE recommends metering, including where electrical-chain losses matter, and discusses retaining measurements for at least one year to obtain annual energy totals. That monitoring guidance does not set the length of the hypothetical ownership-cost period.

When does this stop being an on-premises refresh comparison?

If one option moves the workload to cloud or colocation, redraw the cost boundary. The same delivered service and period still matter, but ownership and responsibility for equipment, space, power, cooling and operations differ.

In a traditional colocation arrangement, a customer can own and manage its IT equipment while renting facility services; management services may also be offered. A cloud arrangement assigns infrastructure operations differently. Compare the responsibilities and charges in the proposed contract rather than placing a service fee into the on-premises equipment column.

What would make a cost decision possible?

First confirm that both options meet the specified service and continuity requirements. Then fill the matrix with acquisition and deployment quotes, likely intervention dates and costs, credible remaining-life assumptions, load-based energy measurements or supported estimates, applicable electricity terms and the facility changes each option would actually cause. Keep unresolved entries visible rather than replacing them with assumed savings.

The general guidance cannot rank these options or establish a break-even point. Its useful implication is narrower: once common costs are separated from costs caused by each choice, the missing measurements and quotations become clear—and only then can a project-specific total-cost comparison support a decision.

Sources

• IBM — What Is a Data Center? | IBM
• IBM — What Is a Green Data Center? | IBM
• U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

Related reading

• Can You Compare Data Center Electricity Use by PUE Alone?
• Who Operates and Uses Data Centers? Ownership and Responsibility Models

Long-Haul DCI Fiber Testing: When Is Characterization Needed Beyond Insertion Loss?

A long-haul DCI label does not, by itself, justify ordering chromatic-dispersion (CD), polarization-mode-dispersion (PMD), or spectral-attenuation tests.

First identify the transmission system and installed end-to-end path. This guide separates what a loss test proves from what other measurements can reveal, then uses a decision matrix to set a link-specific test scope.

Data center deep-dive series

What must be known before setting the test scope?

Start with the equipment and the path it will use: optical interface and data rate, operating wavelengths and any CWDM or DWDM plan, route length, fiber types and segments, connectors, and splices. Those details support the link’s loss budget and show which other optical properties might matter.

If the equipment’s limits or the end-to-end route are still unknown, additional tests and acceptance values cannot yet be specified responsibly.

A new route being assessed for a specified future high-speed system and an older route being upgraded to specified equipment are both reasons to review characterization—but neither calls for an automatic test bundle. The Fiber Optic Association (FOA) describes dispersion as a possible concern on long, high-speed links above 2.5 Gb/s.

That figure is a prompt to investigate, not a threshold requiring field CD and PMD tests on every DCI link above it.

The Fiber Optic Association — The FOA Reference For Fiber Optics - Fiber Optic Network Design

The Fiber Optic Association — The FOA Reference For Fiber Optics - Testing - Chromatic Dispersion and Polarization Mode Dispersion

What does an insertion-loss PASS leave unanswered?

For installed cable-plant loss acceptance, measure end-to-end insertion loss with a light source and power meter or an optical loss test set (OLTS), then compare the result with that path’s loss budget. Record the test wavelength and the method used to establish the 0 dB reference; otherwise, the measured loss is difficult to interpret consistently.

A PASS establishes the loss result at the tested wavelength under the stated conditions. It does not establish acceptable CD, PMD, or attenuation across an entire planned wavelength band.

Other measurements answer other questions: reflectance or optical return loss (ORL) addresses reflection concerns, while an optical time-domain reflectometer (OTDR) can help locate splice or installation-stress problems.

A long outside-plant route with splices gives a reason to consider OTDR testing, but an OTDR trace alone does not replace end-to-end insertion-loss acceptance. Nor should the same OTDR scope be imposed automatically on a short premises link.

The Fiber Optic Association — The FOA Reference For Fiber Optics - Testing Cables

When do CD, PMD, and spectral attenuation enter the decision?

For CD, estimate the end-to-end dispersion from each fiber segment’s dispersion coefficient at the operating wavelength and its length, then compare the estimate with the specified system’s tolerance. FOA’s illustrative calculation uses 17 ps/nm/km for G.652 fiber at 1550 nm: over 50 km, that gives 850 ps/nm.

It is an estimate for those stated inputs, not a measured result or a universal DCI acceptance limit. A route assembled from different fibers cannot be judged from one fiber type’s nominal value alone.

If PMD could limit the chosen system, check the equipment’s tolerance and specify an evaluation appropriate to the installed route. PMD is affected by fiber characteristics and stress; differential group delay can vary with wavelength and time.

A reported delay and a statistically specified PMD coefficient are not interchangeable, and isolated measurements made at different times or by different methods need careful interpretation.

If a CWDM or DWDM plan uses multiple wavelengths, consider spectral attenuation over the bands the system will actually use—particularly when an older fiber’s wavelength-dependent performance is unknown. A loss PASS at one wavelength cannot establish performance throughout those bands. This does not mean every WDM path needs the same full-band test.

When ordering a characterization test, name both the quantity and the measurement method. FOA describes CD methods that measure across multiple wavelengths, including methods requiring access to both fiber ends and a specialized single-ended OTDR-based method. An ordinary OTDR splice-inspection trace is not, merely by being an OTDR trace, a CD or PMD result.

Does coherent transmission change the answer?

Yes, but it does not settle it on its own. Digital signal processing in a coherent receiver can compensate for some CD- and PMD-related degradation, so a test decision made for a noncoherent long-haul system should not simply be copied over.

Compensation also has limits: calling a link “coherent” does not establish that its installed path, especially one containing older fiber, is suitable without further evaluation. Compare the specified equipment’s tolerances and required cable-plant tests with the path’s known characteristics.

The Fiber Optic Association — The FOA Reference For Fiber Optics - Coherent Fiber Optic Data Links -

What should the link-specific test plan say?

Use these branches together where more than one applies. They identify a decision to make, not universal DCI test classes.

Equipment limits, operating wavelengths, or end-to-end route are unknown

Scope decision

Defer the additional-test scope and its acceptance values

What to specify or record

Missing equipment and path information

Specified equipment has no additional requirement, and review of the route identifies no separate characterization concern

Scope decision

Perform insertion-loss acceptance without adding a blanket characterization suite

What to specify or record

Basis for exclusion, loss budget, test wavelength, and 0 dB reference method

Long outside-plant path has splices, or reflection is a concern

Scope decision

Consider OTDR or reflectance/ORL testing for the particular diagnostic question; retain separate end-to-end loss testing

What to specify or record

Splice or fault-location question, or applicable reflection condition

New path has a specified future high-speed use, or an existing path will be upgraded to specified equipment

Scope decision

Compare system tolerances with the route; select CD or PMD evaluation where needed

What to specify or record

Operating wavelength, route-based estimate or evaluation method, and equipment limit

WDM operating band is defined, especially on older fiber with unknown spectral behavior

Scope decision

Assess whether spectral-attenuation testing is needed over the intended band

What to specify or record

Wavelength band and applicable acceptance condition

Coherent system is specified

Scope decision

Neither mandate nor waive characterization solely because DSP is present

What to specify or record

Equipment’s required tests and tolerances, checked against path information

For each test selected, tie the method, wavelength or band, acceptance value, and result to the identified end-to-end path and fiber. If an acceptance value is not yet available, record it as unresolved rather than inventing one. That distinction keeps today’s loss acceptance from being mistaken for proof that the same fiber will support a later transmission system.

Sources

• The Fiber Optic Association — The FOA Reference For Fiber Optics - Fiber Optic Network Design
• The Fiber Optic Association — The FOA Reference For Fiber Optics - Testing - Chromatic Dispersion and Polarization Mode Dispersion
• The Fiber Optic Association — The FOA Reference For Fiber Optics - Testing Cables
• The Fiber Optic Association — The FOA Reference For Fiber Optics - Coherent Fiber Optic Data Links -

Related reading

• Connecting Two Data Centers: A Hypothetical DCI Design Example
• Fiber Cable-Plant Testing vs. Application Acceptance: Why a Loss PASS Is Not a Service PASS

How to Evaluate a Data Center UPS: Define the Protected Load and Power Path First

A UPS rating does not tell you which equipment is protected.

Use a decision matrix to separate the IT loads that need protection, their actual electrical connections, and the operating conditions behind an efficiency claim. Leave site-specific capacity, runtime and transfer performance unverified until the relevant records are available.

Data center deep-dive series

What must be decided before comparing UPS ratings?

Start with the equipment that needs backup power, then trace each of its power inputs through the proposed electrical distribution system. Only after that boundary is clear can you compare efficiency under the loads the system is expected to carry. The U.S.

Department of Energy’s Best Practices Guide for Energy-Efficient Data Center Design recommends first assessing whether all equipment—or only part of a data center—requires a UPS to reduce UPS-related losses. A rating alone cannot answer that question.

U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

How do you record what is protected—and what is not?

Servers, storage and network equipment are useful categories for the protected-load list, not a rule that every device must be connected to a UPS. DOE notes that the share of IT power needing UPS protection can differ substantially between, for example, a scientific computing facility and a financial institution.

Use “excluded” only where the protection requirement has been established; use “unverified” where the requirement or connection is unknown.

Protected IT load

Record for each proposal

Identify which server, storage and network equipment requires UPS protection, and why.

Exclude or defer when

Exclude equipment only if it has been determined not to require that protection; otherwise leave its status unverified.

Power inputs and path

Record for each proposal

Trace each protected input through its actual distribution connections. Identify the relevant utility service, switchboard or switchgear, alternate source, UPS and PDU, as applicable.

Exclude or defer when

Do not infer that both inputs of a dual-power-supply server are protected. Defer that claim until each input’s connection is confirmed.

Backup arrangement

Record for each proposal

Identify the proposed UPS technology, the generator’s role if one is present, and the required backup duration.

Exclude or defer when

Do not infer a battery system or a particular runtime from the word “UPS.”

Operating load

Record for each proposal

Record initial and future loads, including expected part-load or low-load operation and the redundancy arrangement.

Exclude or defer when

Do not rank proposals by a single rated-load efficiency figure when their operating conditions differ.

Efficiency boundary

Record for each proposal

State the UPS operating mode and whether the figure covers the UPS alone or also includes downstream distribution losses.

Exclude or defer when

Do not treat figures with different equipment boundaries as directly comparable.

Decision evidence

Record for each proposal

Check site load and connection records against equipment specifications and test evidence for capacity, runtime and transfer performance.

Exclude or defer when

Leave missing site-specific values unverified rather than declaring the proposal suitable or unsuitable.

This is a list of components to trace, not a fixed wiring sequence. DOE describes these as elements of a typical distribution system and explains that a PDU can distribute power received from a UPS or generator to multiple devices. UPS backup technology may use batteries, rotary machines, fuel cells or other approaches.

IBM’s Data centers overview also describes generators as a possible response to more severe outages. Neither description establishes how a particular site is connected.

IBM — What Is a Data Center? | IBM

When are two efficiency figures comparable?

Compare them at a defined protected load, expected part-load condition, redundancy state and operating mode. DOE advises considering initial and future loads—particularly part-load and low-load conditions—when selecting electrical equipment. Redundancy can change the load carried by each UPS even when the protected load stays the same.

In a DOE teaching example of battery-based N+1 systems, two UPS units each operate at 30% load factor; serving the same load with three smaller units puts each at 40%. The example illustrates a possible efficiency difference under those conditions, not a guarantee that three units are better at every site.

Mode and equipment boundary matter too. DOE describes a more efficient line-conditioning mode available on some double-conversion UPS systems; it should not be assumed to exist on every product or to provide the same power-conditioning conditions as double-conversion operation.

Power is also lost within the UPS, while a PDU with a built-in transformer can lose energy as heat during voltage conversion. Transformer load-factor guidance applies to configurations that actually contain that transformer, not to every PDU.

Keep facility PUE separate from UPS efficiency. In DOE’s definition, site-power PUE is the facility’s total annual energy divided by the annual energy drawn by all IT equipment. It helps characterize supporting infrastructure; it does not report the efficiency of one UPS or the useful work performed by IT equipment.

A proposal comparison should therefore identify both the operating conditions and which losses its efficiency figure includes.

What evidence is needed for a final choice?

Plan how a claimed efficiency could be checked after installation. DOE describes more detailed monitoring of losses along the electrical path, including transformers, UPS equipment and PDUs with transformers. It advises specifying meter accuracy and calibration status, trending measurements, and retaining data long enough to obtain annual energy totals.

The measurement points must match the equipment boundary used in the comparison.

Efficiency cannot substitute for protection. DOE notes that reliability and the ability to accommodate high power density can take priority over energy efficiency for critical data center loads.

The useful decision order is to establish the protected load, verify each input’s actual path, compare losses under equivalent operating conditions, and leave unsupported values open. Without site load and wiring records and the proposed equipment’s specifications and test evidence, capacity, backup duration and transfer performance remain questions to resolve—not conclusions supplied by a general design guide.

Sources

• IBM — What Is a Data Center? | IBM
• U.S. Department of Energy — Best Practices Guide for Energy-Efficient Data Center Design

Related reading

• How Does Power Reach IT Equipment in a Data Center?
• Can You Compare Data Center Electricity Use by PUE Alone?