CRAC Unit Sizing for Data Rooms: Summing Sensible Heat Loads, Sizing Margin, and Sensible-versus-Total Capacity

Why a Data Room Is a Sensible-Heat Problem, Not a Comfort-Cooling One

A data room is cooled differently from an office because almost all of its heat is sensible: servers convert electrical power into dry heat with essentially no moisture, so a CRAC unit is a sensible-cooling machine sized to remove that dry heat and hold the rack inlets within the environmental envelope, not a comfort unit balancing temperature and humidity for people.

In an office, cooling handles both sensible heat (warm air) and latent heat (moisture from people and outdoor air), so equipment carries a sensible heat ratio (SHR) around 0.7 to 0.8. In a data room, the people are few and the outdoor air is minimal, so the load is almost pure sensible, SHR near 0.9 to 1.0. The IT equipment turns nearly all its electrical draw into heat that must be removed at the same rate it is produced, continuously. That is why data-room cooling is sized on sensible heat and why using a comfort-cooling total capacity figure misleads.

The calculator sums the room's sensible heat sources (IT equipment, lighting, people, envelope gains, ventilation/infiltration), applies a design margin, and returns a recommended CRAC size in tons, kW, and BTU/hr. It is a first-pass selection screen, opening the data-center design workflow. The result feeds the later questions this cluster covers: redundancy (how many units), containment (how the air is delivered), raised-floor pressure (how it is distributed), and efficiency (PUE). ASHRAE TC 9.9 defines the environmental envelope the cooling ultimately serves; TIA-942-C frames the infrastructure. This is where the sizing starts.

Calculator Inputs: The Five Room Loads, Margin, and a Selected Unit

The calculator accepts eight inputs across two unit systems and returns the recommended CRAC capacity with a classification badge.

Unit system. Imperial (BTU/hr, tons) or Metric (kW). The calculation logic is identical; only the display changes.

IT Equipment Load (kW or BTU/hr). Heat from servers, switches, storage, UPS losses. The dominant source at 70 to 90 percent of total room load. Per ASHRAE Fundamentals, all IT electrical power converts to heat: X kW in = X kW of heat out. Use nameplate power draw (derated for actual utilization) or measured power from a PDU.

Lighting Load (kW or BTU/hr). Heat from room fixtures. Small but real; LED retrofits lower it.

Occupancy Load (kW or BTU/hr). Heat from people. Per ASHRAE Fundamentals, sensible gain per person is approximately 250 to 340 BTU/hr (73 to 100 W). Data rooms are lightly staffed, so this term is small.

Envelope / Miscellaneous Load (kW or BTU/hr). Conduction through walls, ceiling, and floor, plus miscellaneous gains. Small for interior rooms; larger for exterior walls or roof exposure.

Ventilation / Infiltration Load (kW or BTU/hr). Heat from outdoor air entering by code-minimum mechanical ventilation plus door and leak infiltration. Brings both sensible and, in humid climates, latent load (latent is handled separately by dedicated systems).

Sizing Margin. A dropdown from 0 percent (exact) through 5, 10, 15, 20, 25, and 30 percent. The buffer above the required load for growth, uncertainty, and derating. Ten percent is the standard default.

Selected CRAC Capacity (optional, kW or BTU/hr). A candidate unit to check against the load. The calculator returns a sizing margin and classification badge.

Outputs: Required Cooling Load (sum of five), Recommended CRAC Capacity (load times (1 + margin)), size in tons/kW/BTU/hr, and a classification badge from UNDERSIZED through SIGNIFICANTLY OVERSIZED. Conversion factors: 1 ton = 12,000 BTU/hr; 1 kW = 3,412.142 BTU/hr.

The calculator does not account for rack-by-rack airflow, raised-floor pressure balance, hot/cold-aisle containment, humidity-control energy, coil SHR behavior, psychrometrics, redundancy sequencing, chilled-water plant interaction, or the sensible/total split of the selected unit. It is a screening tool, not a mechanical design.

The Five Heat Sources: Why IT Is 70 to 90 Percent and the Rest Still Count

The required cooling load is the sum of five sensible heat sources, and while IT equipment dominates at 70 to 90 percent, sizing on IT alone understates the load enough to matter.

Required Cooling Load = IT + Lighting + Occupancy + Envelope/Misc + Ventilation/Infiltration

IT Equipment (70 to 90 percent). Servers, switches, storage, UPS losses. Nearly all IT electrical power becomes heat: energy in equals heat out. The anchor of the load and the first number to pin down accurately.

Lighting (2 to 5 percent). Fixtures above and within the space. Small, but a 1,000 BTU/hr (0.29 kW) fixture in a 24-hour room adds up over a year's runtime.

Occupancy (1 to 3 percent). Per ASHRAE Fundamentals, sensible heat per seated person is approximately 250 to 340 BTU/hr (73 to 100 W). Data rooms are lightly staffed; a two-person operations team adds roughly 600 BTU/hr (0.18 kW).

Envelope / Miscellaneous (2 to 8 percent). Conduction through the room shell. Small for an interior room with conditioned space on all sides; larger with exterior walls, roof exposure, or sun load penetrating glazing.

Ventilation / Infiltration (3 to 10 percent). Outdoor air brought in by code-minimum ventilation plus door and gap infiltration. In humid climates, latent load from this source is handled by dedicated dehumidification outside the CRAC's scope.

Per ASHRAE Fundamentals, all IT electrical power converts to heat at the same rate. A 7 kW (23,900 BTU/hr) server load means 7 kW of heat to remove continuously. UPS and PDU conversion losses add to it; typical UPS efficiency is 92 to 96 percent, so a 20 kW IT load may draw 21 to 22 kW from the utility, all of which enters the room as heat.

Why the remaining four sources still count is arithmetic. Per Schneider Electric data-center cooling guidance, sizing on IT alone when IT is 80 percent of the total understates the load by 25 percent (100 divided by 80 = 1.25 ratio). On a 30,000 BTU/hr (8.79 kW) room, that is 6,000 BTU/hr (1.76 kW) of unaccounted heat before any margin is applied.

Example proportions:
IT = 24,000 of 30,000 total = 80%
Other four = 6,000 = 20%
Sizing on IT alone (24,000) vs total (30,000): 20% short before margin

Per ASHRAE Fundamentals and Schneider Electric guidance: required load is the sum of five sensible sources. IT dominates (70 to 90 percent), but lighting, people, envelope, and ventilation add 10 to 30 percent. IT-only sizing undersizes before the margin is even applied.

Summing the Load and Applying the Sizing Margin

Once the five loads are summed, the recommended CRAC capacity is that total multiplied by one plus the sizing margin, a deliberate buffer for load growth, measurement uncertainty, and derating.

Required Cooling Load = IT + Lighting + Occupancy + Envelope/Misc + Ventilation/Infiltration

Recommended CRAC Capacity = Required Load × (1 + margin)
  margin = decimal (10% = 0.10)

Unit conversions:

Tons = BTU/hr ÷ 12,000
kW   = BTU/hr ÷ 3,412.142
BTU/hr = kW × 3,412.142

Why a margin is necessary. Load growth: IT loads rise over a room's life as server density increases. Measurement uncertainty: nameplate power overstates actual draw in many configurations; real measured data is preferable but not always available. Derating: a unit's rated sensible capacity drops at higher return-air temperatures, higher altitude, and as coils foul over time. Per standard practice, 10 percent is the common default; 0 to 15 percent is the well-sized band.

The margin is not redundancy. The sizing margin buffers the load estimate; redundancy (Section 10) adds standby units for failure coverage. Confusing them leads to under-provisioning both.

Worked example:

Required = 24,000 + 2,000 + 800 + 1,200 + 2,000 = 30,000 BTU/hr (8.79 kW)
Recommended = 30,000 × 1.10 = 33,000 BTU/hr (9.67 kW)
Tons = 33,000 / 12,000 = 2.75 tons
kW   = 33,000 / 3,412.142 = 9.67 kW

Real CRAC units come in nominal sizes; the recommended capacity shortlists the next available nominal unit at or above it, then checked for sensible capacity (see Section 6).

Sensible versus Total Capacity: The CRAC Rating Trap

A CRAC unit's nameplate often lists total cooling capacity, but a data room needs sensible capacity, and matching the room's sensible load to a unit's total rating oversizes on paper while undersizing the sensible cooling that actually matters.

Total capacity = sensible + latent
SHR (sensible heat ratio) = sensible / total

Data room load: ~pure sensible (SHR 0.9 to 1.0)
CRAC unit: sensible capacity < total capacity (unit SHR < 1.0)

The trap:

Room needs 30,000 BTU/hr SENSIBLE.
A unit rated 33,000 BTU/hr TOTAL may deliver only ~28,000 sensible (SHR 0.85).
Matching room-sensible to unit-total leaves the room UNDER-cooled on sensible.

Why a CRAC's sensible capacity is less than its total: the unit's total rating includes latent (dehumidification) capacity that a dry data room does not need. The sensible portion, the part that removes the dry server heat, is less than the total nameplate. Data-center CRAC units are designed for high SHR (0.9 to 1.0) precisely to maximize sensible cooling, but the rating basis still matters for the comparison.

The correct comparison: match the room's sensible load to the unit's sensible capacity at the design return-air temperature, not to its total nameplate. Manufacturer capacity tables give sensible capacity at specified return conditions. A higher return temperature (warmer hot aisle) raises a unit's sensible capacity; the table lookup must reflect the actual design condition.

This calculator sizes the required load and recommended capacity. It does not split the selected unit's sensible-versus-total. That check is the engineer's next step against manufacturer data.

Correct selection: unit whose SENSIBLE capacity ≥ 30,000 BTU/hr at design return temperature.
Not: unit whose TOTAL nameplate ≥ 33,000 BTU/hr.

Per ASHRAE Fundamentals and manufacturer practice: data rooms need sensible capacity; CRAC total nameplate includes latent the room does not use. Match the room's sensible load to the unit's sensible capacity at design return conditions, not to total capacity.

The Sizing Margin Framework: Undersized to Significantly Oversized

The calculator classifies a selected unit against the required load on a fixed margin scale, and knowing the bands guides whether a candidate is a good match.

Sizing Margin vs Selected (%) = ((Selected − Required) / Required) × 100

Margin vs required	Classification
Below 0%	UNDERSIZED
0% to 15%	WELL SIZED
15% to 30%	SLIGHTLY OVERSIZED
Above 30%	SIGNIFICANTLY OVERSIZED

Undersized (below 0 percent). The unit is smaller than the load. The room cannot hold temperature at peak. Reject, unless redundant units provide the balance.

Well sized (0 to 15 percent). The target band. Enough buffer without waste. Stable, efficient, steady compressor operation.

Slightly oversized (15 to 30 percent). Acceptable for variable-capacity units (modulating compressors, EC fans, digital scroll) that turn down. Risky for fixed-speed units that cannot reduce output.

Significantly oversized (above 30 percent). Short-cycling risk, poor humidity control, wasted capital. Avoid unless units are staged modularly or modulation is confirmed.

Worked example:

Selected 36,000 vs required 30,000: (36,000 − 30,000) / 30,000 × 100 = 20% → SLIGHTLY OVERSIZED
Acceptable for a variable-capacity unit; short-cycling risk for fixed-speed.

Per standard practice: the well-sized band is 0 to 15 percent; 15 to 30 percent slightly oversized (fine for modulating units); above 30 percent significantly oversized with short-cycle risk. Band interpretation depends on whether the unit has capacity modulation.

Why Oversizing a CRAC Is Its Own Failure Mode

The instinct to oversize for safety backfires in a data room, because an oversized fixed-speed CRAC short-cycles, and short-cycling degrades humidity control, efficiency, and compressor life.

The oversizing cascade:

1. Oversized unit satisfies the sensible load quickly
2. Compressor reaches setpoint, shuts off (short cycle)
3. Frequent start/stop cycling resumes
4. Each start draws inrush current, thermally stresses the compressor
5. Short run times prevent stable humidity stabilization
6. Room humidity swings; efficiency drops

Humidity control suffers. A CRAC controls humidity during steady run. Short cycles never run long enough to stabilize humidity, so the room drifts, sometimes causing adjacent units to fight each other (one humidifying while another dehumidifies).

Compressor life shortens. Frequent cycling is the enemy of compressor longevity. Inrush current and thermal cycling wear the machine faster than steady operation at part load.

Efficiency drops. Cycling losses (startup transients, off-cycle heat exchange) lower delivered efficiency versus steady modulation at design conditions.

Capital is wasted. Oversized units cost more to buy and install for capacity that is never used and actively harms operation.

Why data rooms are sensitive: the IT load is steady (servers run 24/7), so a right-sized unit runs steadily. Oversizing breaks that natural match, introducing cycling where none is needed.

The variable-capacity exception: modulating units (EC fans, digital scroll, variable-speed drives) absorb oversizing by turning down, avoiding the cycling penalty. Per ASHRAE and vendor guidance, fixed-speed units do not share this tolerance.

ASHRAE TC 9.9 and the Rack-Inlet Target the Cooling Serves

The point of CRAC sizing is not simply to remove heat but to hold the server rack inlets within the ASHRAE TC 9.9 environmental envelope, and that rack-inlet target, not the room average, is the real design condition.

ASHRAE TC 9.9 (Thermal Guidelines for Data Processing Environments) defines:

Recommended envelope: 18 to 27°C (64.4 to 80.6°F) rack inlet
Allowable envelopes (A1 through A4): wider ranges, up to 32 to 45°C depending on class
Dew-point and humidity limits: also defined per class

The temperature that matters is the air entering the server at the rack inlet in the cold aisle, not the room average. A room can measure an acceptable average temperature while specific rack inlets in high-density zones exceed the envelope. Sizing provides the capacity; airflow management (containment, raised floor) delivers it to the inlets.

Sufficient total capacity + poor air delivery = hot rack inlets from recirculation
The heat is removed on average, but specific inlets exceed TC 9.9 limits

ASHRAE Standard 90.4 (Energy Standard for Data Centers) ties efficient operation to holding rack-inlet conditions within TC 9.9, not to nameplate tonnage. TIA-942-C frames the infrastructure around maintaining those conditions at the specified tier.

Operating warmer within the recommended or allowable envelope saves energy: higher return temperature raises a CRAC unit's sensible capacity and extends air-side economizer hours. Modern practice routinely runs at 24 to 27°C (75 to 81°F) rack inlet rather than legacy 20°C (68°F) setpoints, materially lowering cooling energy per ASHRAE 90.4.

Per ASHRAE TC 9.9 and ASHRAE 90.4: CRAC sizing serves the rack-inlet envelope (recommended 18 to 27°C, 64 to 81°F), not the room average. Capacity must be paired with air delivery to hold every rack inlet. Warmer inlets within the envelope raise capacity and lower energy.

Redundancy: Why N+1 and 2N Multiply the Installed Capacity

The required load sizes the cooling the room needs, but the installed CRAC capacity is larger, because redundancy adds standby units so cooling survives a unit failure or maintenance.

Redundancy configurations:

N:   exactly the capacity needed, no spare (no redundancy)
N+1: one extra unit beyond need (survives one failure)
N+2: two extra units (survives two failures)
2N:  full duplicate system (two independent full-capacity systems)

Why redundancy: IT runs continuously. A cooling failure overheats servers in minutes. Per Uptime Institute Tier Standard, redundancy keeps cooling available during a unit failure or planned maintenance without taking the load offline.

Installed capacity math:

Required = 30,000 BTU/hr. One CRAC = 33,000 BTU/hr (2.75 tons).

N:   one unit — no spare
N+1: two units (one running, one standby) → 66,000 BTU/hr installed
2N:  two full independent systems → 132,000 BTU/hr installed

Uptime Institute Tier Standard sets the configuration by tier:

Tier I to II:  N or N+1 (basic)
Tier III:      N+1 concurrently maintainable
Tier IV:       2N fault tolerant

With multiple units sharing the load, each runs at part load; a failure shifts the load to the survivors, which must carry the full sensible load at design inlet conditions. Verify single-unit sensible capacity against the full load, not part-load conditions.

This calculator sizes the load for one room and returns required and recommended capacity. The redundancy overlay, how many units and which tier, is the engineer's next decision, multiplying installed capacity above the room load.

Per Uptime Institute Tier Standard and TIA-942-C: redundancy (N+1, N+2, 2N) adds standby units so cooling survives failure and maintenance, multiplying installed capacity above the room load. The tier sets the configuration. Redundancy is separate from the sizing margin.

Worked Example: A 30,000 BTU/hr Data Room to a 2.75-Ton CRAC

Scenario: small data room, five loads, 10 percent margin.

Step 1. The five loads:

IT Equipment          = 24,000 BTU/hr  (7.03 kW)
Lighting              =  2,000 BTU/hr  (0.59 kW)
Occupancy             =    800 BTU/hr  (0.23 kW)
Envelope / Misc       =  1,200 BTU/hr  (0.35 kW)
Ventilation / Infilt. =  2,000 BTU/hr  (0.59 kW)

Step 2. Required cooling load:

Required = 24,000 + 2,000 + 800 + 1,200 + 2,000 = 30,000 BTU/hr (8.79 kW)

Step 3. Load composition:

IT = 24,000 / 30,000 = 80% (dominant)
Other four = 6,000 = 20%

Step 4. Apply 10 percent sizing margin:

Recommended = 30,000 × 1.10 = 33,000 BTU/hr (9.67 kW)

Step 5. Convert to tons:

Tons = 33,000 / 12,000 = 2.75 tons

Step 6. Convert to kW:

kW = 33,000 / 3,412.142 = 9.67 kW

Step 7. Sensible capacity check (the real selection step):

Select a nominal CRAC unit whose SENSIBLE capacity ≥ 30,000 BTU/hr at design return temperature.
A 3-ton nominal unit (36,000 total) may deliver ~30,000 to 33,000 BTU/hr sensible at high SHR: verify against manufacturer data.
Do NOT assume a 33,000 total-nameplate unit gives 30,000 BTU/hr sensible.

Step 8. Nominal selection:

Next nominal size at or above 2.75 tons: typically a 3-ton unit.
Confirm its SENSIBLE capacity meets 30,000 BTU/hr at the design rack-inlet/return condition.

Step 9. Result and next steps:

Recommended: ~2.75 tons (33,000 BTU/hr, 9.67 kW)
Select a 3-ton unit by SENSIBLE capacity.
Then overlay redundancy (N+1 doubles installed), airflow delivery (containment), and rack-inlet verification (TC 9.9).
This is a screening result, not a final mechanical design.

IT is 80 percent of the load; the other 20 percent (lighting, people, envelope, ventilation) is not negligible. Cross-reference the data-center siblings: redundancy sets how many units, containment sets air delivery, raised floor sets distribution.

Selected-Unit Check and the Redundancy Overlay

Selected-unit check:

Selected CRAC = 36,000 BTU/hr
Required load = 30,000 BTU/hr

Margin vs selected:

Sizing Margin = ((36,000 − 30,000) / 30,000) × 100 = 20%

Classification:

20% falls in the 15 to 30% band → SLIGHTLY OVERSIZED

For a variable-capacity unit (modulating compressor or EC fan), 20 percent over the load is acceptable; the unit turns down to match. For a fixed-speed compressor, 20 percent invites short-cycling; prefer staying closer to 10 percent.

Sensible caveat: the 36,000 BTU/hr is the total nameplate. Check the unit's sensible capacity against the 30,000 BTU/hr sensible load at design return conditions. A 20 percent total margin may be a smaller sensible margin.

Redundancy overlay:

Required 30,000 BTU/hr. One unit: 33,000 BTU/hr (2.75 tons).

N+1: two units → 66,000 BTU/hr installed; one can fail
2N:  two full systems → 132,000 BTU/hr installed

With N+1 and both units running, each carries approximately 15,000 BTU/hr (half load). On failure, the survivor must carry the full 30,000 BTU/hr sensible at design inlet conditions. Verify that single-unit sensible capacity covers the full load before specifying N+1 with equal-size units.

Tier III (N+1 concurrently maintainable) vs Tier IV (2N fault tolerant) sets the count.
Margin and redundancy stack: 10% sizing margin AND N+1 redundancy are separate.
Margin buffers the load estimate; redundancy covers failure. Both apply simultaneously.

Per Uptime Institute Tier Standard and standard practice: the selected unit at 36,000 BTU/hr is 20 percent slightly oversized (fine for modulating units). N+1 doubles installed capacity; each surviving unit must carry the full sensible load on failure. Check the unit's sensible capacity, not only its total nameplate.

Application Boundaries: Airflow, Containment, Humidity, Chilled-Water Plant

The calculator applies to: room-level sensible load summation plus margin, recommended CRAC capacity in tons/kW/BTU/hr, and first-pass candidate selection. The following require separate analysis.

Rack-by-rack airflow. The calculator gives room capacity, not airflow distribution to each rack. Hot spots at specific rack inlets need CFD or airflow analysis; total capacity alone does not guarantee inlet temperatures within TC 9.9.

Hot-aisle/cold-aisle containment. Containment changes effective cooling delivery by reducing recirculation and raising return-air temperature, which increases sensible capacity. Its effect on system performance is a separate analysis (covered in the data-center cluster sibling on containment efficiency).

Raised-floor pressure. Underfloor plenum pressure balance and tile placement set how air reaches the cold aisle. A separate calculation (cluster sibling) governs this.

Humidity control. The dry-room assumption holds for most data rooms, but humidity-control energy, humidification/dehumidification interaction, and dew-point limits per TC 9.9 are separate. CRAC units fighting each other on humidity is a real operational issue in rooms with mixed unit control strategies.

Sensible-versus-total of the selected unit. The calculator does not split the unit's sensible/total capacity. That check is against manufacturer data at design conditions.

Redundancy sequencing and staging. How multiple units stage, share load, and sequence on failure is a controls and design task beyond load sizing.

Chilled-water CRAH plant. For chilled-water air-handler units, the chiller, pump, and cooling-tower sizing and interaction are separate from the room load screen.

Part-load and economizer. Annual energy, part-load performance, and air-side or water-side economizer hours require energy modeling per ASHRAE 90.4. The room-load calculator does not project annual energy.

Latent and outdoor air. Significant outdoor-air fractions or humid climates add latent load handled by dedicated systems, separate from the sensible CRAC.

Per ASHRAE TC 9.9, ASHRAE 90.4, and TIA-942-C: room-level sensible sizing is the calculator scope. Airflow distribution, containment, raised-floor pressure, humidity energy, redundancy sequencing, chilled-water plant, and economizer energy require separate analysis. A qualified data-center mechanical engineer prepares the final design; rack-inlet conditions govern compliance.

CRAC Unit Sizing Calculator

FAQ

Related Calculators

• Data Center CRAC Redundancy Calculator: The redundancy overlay (N+1, 2N) that multiplies the installed capacity this sizing produces.
• Hot Aisle Containment Efficiency Calculator: How containment delivers the cooling to the rack inlets, raising effective sensible capacity.
• Raised Floor Pressure Drop Calculator: Underfloor plenum pressure that distributes CRAC airflow to the cold aisle.
• Cooling Load Calculator: General space cooling load for non-data-center rooms.
• AC Tonnage Calculator: Convert cooling load to nominal tons for equipment selection.
• Fan Power Calculator: Fan power for the airflow the CRAC unit delivers.
• Air Density Calculator: Air density correction for altitude and temperature in cooling calculations.
• Air Changes Per Hour Calculator: Ventilation air changes for the room.

Standards References

• ASHRAE TC 9.9, Thermal Guidelines for Data Processing Environments, 5th ed. (2021). Defines recommended (18–27°C) and allowable (A1–A4) rack-inlet envelopes, dew-point limits, and humidity bounds.
• ASHRAE Standard 90.4-2019, Energy Standard for Data Centers. Sets energy-efficiency requirements and ties compliance to maintaining TC 9.9 rack-inlet conditions.
• ASHRAE Handbook of Fundamentals, Chapter 18 (Heat, Air, and Moisture Control) and Chapter 26 (Heat Transfer). Sensible/latent load methods, SHR definitions, and conversion factors used throughout.
• TIA-942-C, Telecommunications Infrastructure Standard for Data Centers (2017). Defines tier infrastructure requirements (Tier I–IV) for cooling redundancy, concurrent maintainability, and fault tolerance.
• Uptime Institute, Tier Standard: Topology (2022). Defines N, N+1, N+2, and 2N redundancy configurations and the Tier I–IV framework; governs how many CRAC units a data center installs.
• Schneider Electric, Data Center Projects: System Design (White Paper 14, Rev. 5, 2018). Provides data-center cooling load estimation methodology including the five sensible heat sources and sizing margin guidance.
• Vertiv, Data Center Cooling — CRAC and CRAH Unit Selection Guide (2020). Manufacturer guidance on sensible heat ratio, total-versus-sensible capacity, and unit sizing for data rooms.
• Stulz, CRAC Unit Sizing and Redundancy Planning for Critical Environments (Technical Reference, 2019). Covers SHR, oversizing risk, and redundancy configurations for CRAC/CRAH systems.