Cooling
CRAC / CRAH Sizing Reference
Sensible heat ratio, derating factors, redundancy modes and target utilization guidance for precision air conditioning sizing. Use alongside the CRAC / CRAH Sizing tool.
Last updated: June 2026
Terms
- CRACComputer Room Air Conditioner
- The specialized air-conditioning unit that cools a server room.
- CRAHComputer Room Air Handler
- Similar job to a CRAC unit — it cools a server room — but instead of cooling the air itself, it uses chilled water piped in from elsewhere in the building.
Sensible Heat Ratio (SHR)
CRAC/CRAH nameplate capacity is usually total capacity (sensible + latent). Server rooms have almost no latent (humidity) load, so only the sensible portion is useful for IT cooling — the SHR tells you what fraction that is.
| Room type | Typical SHR | Notes |
|---|---|---|
| Dedicated server/IT room, no occupants | 0.95 – 1.00 | Almost entirely sensible load — IT equipment doesn't add humidity |
| Mixed-use technical room, light staffing | 0.90 – 0.95 | Small latent contribution from people and outside-air infiltration |
| Comfort-cooling-rated unit used for IT space | 0.70 – 0.85 | Standard comfort AC nameplate SHR — oversized for sensible-only IT load if used as-is |
Rule of thumb: a 12.5 kW nameplate unit at SHR 0.95 delivers about 11.9 kW of usable sensible cooling for IT load.
Unit Derating Factors
| Factor | Typical derating | Why |
|---|---|---|
| Dirty/aging filters | 5 – 10 % | Reduced airflow across the coil cuts effective capacity |
| Altitude (1 500 – 2 000 m) | 8 – 12 % | Lower air density reduces both airflow mass and heat-transfer effectiveness |
| High condenser ambient (DX units) | 10 – 20 % | Hot outdoor/plant-room ambient reduces compressor capacity on warm days |
| Combined general allowance (default) | 10 % | Reasonable starting planning value when site-specific data isn't available yet |
Redundancy Modes
| Mode | Units needed for X required | Typical use |
|---|---|---|
| N | Exactly enough units, no spare | Non-critical / lab spaces only — any single failure causes a cooling shortfall |
| N+1 | One additional unit beyond the calculated minimum | Standard for most server/technical rooms — tolerates one unit failure or maintenance |
| 2N | Two fully independent sets, each independently sized for 100 % of load | Mission-critical facilities — tolerates loss of an entire cooling path |
Target Utilization
Running units near 100% of rated capacity leaves no headroom for derating drift, fouling or hot days. Most designs target a working capacity below nameplate.
| Target utilization | When to use |
|---|---|
| 70 % | Conservative — older units, harsh ambient conditions, or limited maintenance access |
| 80 % (common default) | Balanced planning margin for most new installations |
| 90 %+ | Tight retrofit constraints with strong confidence in nameplate data — leaves little margin |
Frequently Asked Questions
What's the difference between CRAC and CRAH?
CRAC (Computer Room Air Conditioner) units have their own integrated direct-expansion (DX) refrigeration circuit and condenser. CRAH (Computer Room Air Handler) units instead use chilled water from a central plant, with no on-board compressor. CRAH units are typically more efficient at scale but depend entirely on the chilled-water plant's reliability.
Why does sensible heat ratio matter so much for IT cooling?
A unit's total nameplate capacity includes both sensible (temperature) and latent (humidity/dehumidification) cooling. Server rooms generate almost no latent load, so only the sensible fraction does useful work. Using total nameplate capacity instead of sensible capacity overstates how much real IT load a unit can handle — sometimes by 15–30%.
How do I choose between N+1 and 2N?
N+1 protects against a single unit failure or planned maintenance — appropriate for most business-critical but not life-safety-critical IT spaces. 2N protects against loss of an entire cooling path (e.g. one chilled-water loop or one electrical distribution branch feeding multiple units) — reserved for mission-critical facilities where even a brief multi-unit outage is unacceptable.
Should I size for today's load or future growth?
Both, separately. Size the immediate requirement accurately using current/near-term load, then add the safety margin and redundancy mode on top. Don't inflate today's load estimate to "leave room for growth" — that distorts the SHR/derating math. Plan a phased capacity-add instead if significant growth is expected.