Key takeaways
- A 5 kW rack at a 10 K temperature rise needs 0.41 m3/s, or about 875 CFM, and at 15 K the same load needs only 585 CFM, so the inlet-to-outlet rise is a design choice with a cost.
- One open 1U space leaks about 140 CFM at a 15 Pa rack differential, so 10U of unblanked space recirculates more air than the 875 CFM the rack is supplied with.
- A blanking panel with a 1 mm gap per side leaks about 7 CFM, which is why 10U of panels cuts recirculation from roughly 1,400 CFM to under 100 CFM.
- A 63 percent open door gives 0.68 m2 of free area on a 600 x 1,800 mm door; a 78 percent door gives 0.84 m2 and drops about 6 Pa of static pressure at 875 CFM.
- Seal the U space before buying containment: containment stops aisle mixing, but it cannot recover air that never entered the rack.
A 5 kW rack in a 10 K cold aisle needs 875 CFM. That single number decides most rack airflow arguments, because the cooling system can only deliver what the rack lets in, and the rack leaks most of it somewhere else.
The arithmetic is the standard heat balance. Air at 1.2 kg/m3 and 1,005 J/kgK carrying 5 kW through a 10 K rise moves 0.41 m3/s, which is 875 CFM. Allow a 15 K rise and the same 5 kW needs 585 CFM. Every kilowatt costs about 175 CFM at 10 K, or 117 CFM at 15 K.
The design question is not how much air the CRAC can supply. It is how much of that supply reaches the top of the rack at the right temperature. An open U space, a door with too little free area, an unsealed cable cut-out and a gap at the containment ceiling all take air that was meant for a server inlet.
How much air does a rack actually need?
Rack airflow is set by the heat load and the allowed temperature rise, not by the rack size. A 42U frame carrying 3 kW needs 525 CFM at 10 K; the same frame at 12 kW needs 2,100 CFM. The rack does not change, the air does.
That is why airflow and density have to be specified together. A 600 mm wide rack at 3 kW is comfortable on a 63 percent perforated door. The same rack at 12 kW is not, because the door has to pass four times the air through the same 0.68 m2 of free area.
Where rack airflow escapes
The biggest hole in a rack is the one left by a decommissioned server. A 1U opening is 44.45 mm high and about 450 mm wide, or 0.02 m2. At a 15 Pa front-to-back differential, air crosses it at roughly 5 m/s before the entry loss, and a discharge coefficient of 0.65 brings that down to about 3.2 m/s. The result is 0.065 m3/s, or 140 CFM, through one open unit.
Ten open units leak about 1,400 CFM. That is more than the 875 CFM the rack is supplied with, which means the rack stops drawing cold air from the floor and starts pulling warm exhaust back through its own front. The inlet temperature then climbs towards the exhaust temperature, and the servers that were fine at 3 kW start to throttle at 5 kW.
A blanking panel closes that path. A 1U panel in 1.2 mm steel with a 1 mm gap all round has a leak area of about 1,050 mm2, roughly one-twentieth of the opening it covers. At the same 15 Pa it passes about 7 CFM, so ten panels leak about 70 CFM instead of 1,400 CFM.
Rack airflow limits by panel, door and seal
The values below are for a 600 mm wide, 42U rack at 5 kW and a 10 K rise, unless the row states otherwise. Static pressure is the front-to-back difference across the rack, not the room pressure.
| Parameter | Typical value | Ganesh limit or note |
|---|---|---|
| Air per kW at 10 K rise | 175 CFM | 117 CFM at 15 K; specify which rise the room is designed for |
| 5 kW rack airflow | 875 CFM (0.41 m3/s) | the number the door, U space and cable entry must pass |
| Open 1U leak at 15 Pa | about 140 CFM | 10U open is about 1,400 CFM, more than the supply |
| Blanking panel leak at 15 Pa | about 7 CFM | 1 mm gap per side on a 1.2 mm panel |
| Blanking panel gauge | 1.2–1.5 mm CRCA | 1.0 mm flexes and the gap opens at the centre |
| Panel fit clearance | 1 mm per side | 2 mm per side roughly doubles leakage |
| Door free area, 63 percent | 0.68 m2 on 600 x 1,800 mm | baseline for racks up to about 8 kW |
| Door free area, 78 percent | 0.84 m2 on 600 x 1,800 mm | about 5 Pa less drop at 875 CFM |
| Door perforation | 5–6 mm holes, 63–78 percent | deburr both faces before coating |
| Door flatness | 1.5 mm/m | a bowed door leaks at the frame seal |
| Cable entry cut-out | 300 x 100 mm | about 207 CFM if left open; brush strip or grommet |
| Containment panel joint | 2 mm maximum gap | gasket or brush seal at every joint |
| Panel finish | 60–80 µm polyester powder | seven-tank pretreatment, 18 ft x 14 ft booth |
Perforated doors: free area is the whole argument
A door is a plate with holes, and its resistance rises with the square of the velocity through those holes. That is why a door that is fine at 5 kW can strangle a rack at 12 kW without any change to the door.
On a 600 x 1,800 mm door, 63 percent open area gives 0.68 m2 of free area. At 875 CFM the face velocity through the perforation is 0.61 m/s and the door drops about 15 Pa. A 78 percent door gives 0.84 m2, drops the velocity to 0.49 m/s and the pressure to about 9 Pa.
At 2,100 CFM, the flow for a 12 kW rack, the same 63 percent door sees 1.46 m/s and about 85 Pa. The 78 percent door sees 1.18 m/s and about 52 Pa. Server fans typically generate 100–200 Pa at the operating point, so an 85 Pa door takes half the available pressure before the air has entered the rack.
Hole size matters less than open area, but it is not irrelevant. Five millimetre holes at 63 percent open area give a more uniform velocity profile than 8 mm holes at the same percentage, and they block a dropped screw better. Deburr both faces: a raised burr on the inside face adds turbulence right where the air turns into the rack.
A door also has to stay flat. A 1,800 mm door in 1.5 mm steel with a single return will bow across its width, and the bow opens a gap at the frame seal that bypasses the perforation entirely. Two returns, or a 2.0 mm skin, keeps the seal closed.
Containment: what it fixes and what it hides
Containment separates supply air from return air so the two do not mix before the rack inlet. It does not increase the air the rack receives, and it does not close a leak inside the rack.
The mixing arithmetic is simple. If 10 percent of the air entering a rack is its own exhaust, the inlet temperature rises by 10 percent of the rack temperature rise. On a 10 K rise that is 1 K; on a 15 K rise it is 1.5 K. The room then has to supply colder air to hold the same server inlet, which costs chiller capacity for a leak that a blanking panel would have closed.
Containment pays when the leak is between aisles rather than inside the rack. A cold aisle with open ends, a missing ceiling panel or an unsealed cable tray above the racks will lose cold air to the room, and the racks at the far end of the aisle run warm. Sealing the aisle fixes that. Sealing the aisle while 10U in every rack is open does very little.
Containment also changes the fire and access case. Doors need to open from inside, panels need to be removable without tools, and the ceiling has to be supported independently of the racks. Those are design requirements, not sheet metal details, and they belong on the containment drawing.
The order of fixes that actually works
- Blank the open U space. Measure the open units, order panels to the actual rack width and fit them with 1 mm clearance. This is the cheapest CFM you will ever recover.
- Seal the cable entries. A 300 x 100 mm open cut-out leaks about 207 CFM at 15 Pa. Brush strips, split grommets or a slotted gland plate close it.
- Check the door free area against the load. Below 60 percent open area, change the door before adding containment. At 12 kW, the difference between 63 and 78 percent is about 33 Pa.
- Seal the aisle. Only after the rack itself is tight does containment stop a leak rather than move it.
- Raise the supply temperature. Once the inlet is stable, the set point can move up, and that is where the energy saving comes from.
How to measure rack airflow before specifying
A vane anemometer and a differential manometer cost less than a containment aisle, and they answer the question the drawing cannot: is the rack being starved, or is it leaking?
Measure the inlet face of each U position, not the door face. A 1U opening is 0.02 m2, so a reading of 1.5 m/s is 0.03 m3/s, or 63 CFM. Take five readings across the width at the top, middle and bottom of the rack and average them. Sum the U readings and compare the total with the heat balance: kilowatts multiplied by 175 CFM at a 10 K rise.
Then measure the differential across the rack, from the cold aisle to the hot aisle. A sealed rack on a 63 percent door sits at 5–15 Pa. Above 25 Pa, the door or the cable entry is the restriction and the fans are working against the enclosure. Below 2 Pa, the rack is so open that the air is going around the equipment rather than through it.
Finally, measure the inlet temperature at the top U and compare it with the supply temperature at the floor grille. A 2 K difference at the top of the rack is normal stratification. A 6 K difference is recirculation, and it will be visible as a warm band on the upper Us before any meter is needed.
Record the three readings on the rack schedule. A supplier who is asked to hold a door open area can work to a number; a supplier who is told the rack "runs hot" can only guess.
What Ganesh runs for panels, doors and containment
Blanking panels, perforated doors, side panels and containment infill are cut on the 12 kW fibre laser with the 40 ft x 16 ft bed, which puts a 2,000 mm door and its full perforation pattern on one sheet in one setup. Forming runs on the 200 T, 5-axis brake with the 4,450 mm bed, so a door return is folded in a single pass and the two ends match. Coating is seven-tank pretreatment into the 18 ft x 14 ft powder booth, with DFT and cross-hatch adhesion recorded per batch.
The perforation is the part that decides the airflow, so the laser program is set for hole diameter and pitch, not for cut speed alone. A 6 mm hole cut at the wrong focus leaves a taper that reduces the effective open area by a few percent, which is invisible on the drawing and measurable at the door.
Panels and doors are built alongside the 19-inch frames they fit, which is why the mounting holes and the panel edge land in the same tolerance chain. The blanking panel design guide covers the panel edge and fixing detail, the perforated door manufacturing guide covers the hole pattern and forming route, and the data centre rack range lists the family. The racks and enclosures hub collects the related guides.
What to put on the drawing
- Rack width, height and the exact U space the panel must cover, in millimetres.
- Panel gauge and edge form: a folded edge is stiffer than a flat edge and seals better.
- Perforation diameter, pitch and open area as a percentage, not just a hole count.
- Door free area at the design airflow, plus the allowable pressure drop.
- Seal detail at the frame: gasket, brush strip or overlap, and the gap it closes.
- Cable entry size and the sealing method.
- Finish, colour, and any masked earthing points.
- Flatness limit, because a bowed door bypasses its own perforation.
When containment is the wrong spend
Containment is the wrong first purchase in a room where the racks are half empty and the open U space has never been blanked. It is also the wrong purchase where the cooling is perimeter CRAC with no raised floor, because there is no plenum to deliver air to the aisle in the first place.
Ganesh does not run CFD, measure rack airflow in a live room or certify fire performance of containment. Those are the consultant's and the fire engineer's scope. What we can do is hold the perforation open area, the panel gap and the door flatness that the airflow calculation assumes.
Send the rack schedule, the door open area and the panel sizes with the enquiry, and the quotation will state the gauge, perforation and seal detail it is based on.
Frequently asked questions
Do blanking panels really save energy?
They save cooling energy by stopping recirculation, not by reducing the load. Ten open U spaces in a 42U rack leak about 1,400 CFM at a 15 Pa differential, which is more than a 5 kW rack needs, so blanking them removes a short circuit that the CRAC is paying to overcome.
What open area should a perforated rack door have?
Use 63 percent as the baseline and 78 percent where the rack carries more than about 8 kW or the door is deeper than 25 mm. Below 60 percent, the door adds more than 15 Pa at 875 CFM and starts to starve the top of the rack.
Is hot aisle or cold aisle containment better?
Cold aisle containment is easier to retrofit and keeps the supply air where the rack inlets are, which suits rooms with a raised floor and perimeter cooling. Hot aisle containment captures the exhaust and suits higher densities, but the ceiling and door seals have to be tighter because the aisle runs warmer.
How much gap is allowed around a blanking panel?
Keep the panel-to-post gap at 1 mm per side or less, and use a folded edge or a brush strip where the panel meets the mounting flange. A 2 mm gap per side roughly doubles the leakage of the panel, and 10U of loose panels then leaks like 2U of open space.
Does a rack need a sealed cable entry?
Yes, if the rack is inside a contained aisle. An open 300 x 100 mm brush-less cable cut-out is 0.03 m2 of free area, which leaks about 207 CFM at 15 Pa, more than the gap around a dozen blanking panels.
Sources and further reading
- ASHRAE Datacom Series — Thermal guidelines for data processing environments · reviewed 3 Nov 2026
- IEC 60529 — Degrees of protection provided by enclosures (IP code) · reviewed 3 Nov 2026
- ISO 2768-2 — General tolerances for geometrical tolerances on features without individual tolerance indications · reviewed 3 Nov 2026
Specifying blanking panels or a perforated door?
Send the rack layout, the door open area and the panel size. We will confirm the gauge, flatness and coating, and quote panels, doors and containment as one build.
Request a quote