Many server problems in small and mid-sized companies start in the room, not in the server: a split air-conditioner that trips at night, a rack with gaps that recirculates hot air, or both power supplies plugged into the same power strip. In Saudi Arabia and the Gulf, where outside temperatures are high for much of the year, the margin for error is small. This guide collects what Dell and HPE actually document about temperature, airflow and power for their rack servers, so you can check a server room against the vendors' own requirements.
Before you start
- The figures below are examples from specific guides (HPE DL360/DL380 Gen11, Dell R760). Always check the user guide or technical guide of your exact model.
- Electrical work must follow local regulations and be done by licensed electricians, as HPE's power requirements state.
- For room-level air-conditioning design, use a proper cooling-load calculation: see cooling load calculation guide.
Step 1 — Keep inlet temperature inside the documented range
| Item | HPE DL360 Gen11 | Dell R760 |
|---|---|---|
| Standard continuous operation | 10°C to 35°C | ASHRAE A2: 10 to 35°C, no direct sunlight |
| Extended ranges | 5 to 10°C, 35 to 40°C and 40 to 45°C only for approved hardware configurations | ASHRAE A3 and A4 allow limited annual runtime above 35°C (10% at 35 to 40°C; A4 adds 1% at 40 to 45°C); check the guide's thermal restriction matrix for your configuration |
| Altitude derating | 1.0°C per 304.8 m up to 3048 m | A2: 1°C per 300 m above 900 m |
| Rate of change | Maximum 20°C per hour | 20°C in an hour and 5°C in 15 minutes |
| Humidity (operating) | Higher of -12°C dew point or 8% RH, up to 24°C dew point or 90% RH | A2: 8% RH with -12°C minimum dew point, up to 80% RH with 21°C maximum dew point |
HPE also states that the maximum recommended ambient operating temperature (TMRA) for most server products is 35°C and that the room temperature must not exceed it. Dell adds an important detail: the supported ambient temperature of a configuration is set by its most limited component. Its example: if the processor and memory support 35°C but the GPU supports 30°C, the whole server is limited to 30°C. Treat 35°C as a ceiling, not a target, and keep your normal room temperature well below it so that one failed air-conditioner does not push servers to their limit.
Step 2 — Get the airflow right
- Front to back: HPE describes servers drawing cool air in through the front of the rack and expelling warm air at the rear. Face all equipment the same way so no server breathes another's exhaust.
- Clearances: for the DL360 Gen11, HPE lists 63.5 cm in front of the rack, 76.2 cm behind it, and 121.9 cm from the back of one rack to the back of another row.
- Rack doors: with a third-party 42U rack that has front and rear doors, HPE requires 5,350 sq cm of evenly distributed holes (64 percent open area) and at least 7 cm side clearance.
- Blanking panels: HPE warns that gaps in the rack change the airflow and that a rack without blanking panels can cause thermal damage. Cover every empty U.
- Blanks inside the server: Dell's R760 manual says drive blanks must be installed in all empty drive slots to maintain cooling, and warns not to run the system without its cover for more than five minutes. HPE says the same for PCIe slots: each needs a cover or a card.
Step 3 — Feed redundant power supplies from separate circuits
Rack servers such as the R760 and DL380 Gen11 take two hot-plug power supplies; Dell documents 1+1 redundancy for the R760. Redundancy only helps if the two supplies are fed from independent sources.
- Connect PSU 1 of every server to circuit or PDU A and PSU 2 to circuit or PDU B, ideally each from its own UPS.
- On Dell, set the policy in iDRAC under Configuration > Power Management > Power Configuration (Power Redundancy Policy). The iDRAC guide explains that with hot spare enabled, one PSU carries most of the load, so if all PSU 2s are on circuit B, circuit B carries much less load and can trigger warnings; with hot spare disabled the load is shared 50-50.
- On HPE, do not mix power supply types in one server: HPE says mixing can limit or disable redundancy and can make the system unstable. Use supplies with the same part number and label colour.
- HPE recommends a PDU that is permanently wired or has a non-detachable cord with an industrial-style plug (for example IEC 60309), and does not recommend common power outlet strips, because of the high ground-leakage current of multiple servers. Never overload the branch circuit that feeds the rack.
Warning: do not test redundancy by pulling power from a production server unless you have a maintenance window and current backups. A wrongly wired rack can lose both feeds at once.
Step 4 — Size the UPS from measured or calculated load
- HPE's power requirements advise protecting servers from power fluctuations and temporary interruptions with a regulating uninterruptible power supply.
- Calculate the load with the vendors' tools rather than PSU nameplate ratings: Dell's technical guides point to the Enterprise Infrastructure Planning Tool (EIPT) at dell.com/calc, and HPE's user guides point to HPE Power Advisor for electrical and heat loading.
- Add every other device on the UPS: switches, firewalls, storage, KVM.
- Choose runtime for a clean shutdown or a generator start, not just for ride-through, and test that the shutdown signal really reaches the hosts.
- Once the servers are running, compare the plan with real readings from iDRAC or iLO power monitoring.
Step 5 — Monitor the room through the servers
Every server already has temperature sensors. Configure iDRAC or iLO alerts so a rising inlet temperature reaches someone before shutdown: HPE's DL360 guide says iLO performs an orderly shutdown when a cautionary temperature level is detected, and the hardware shuts down immediately at a critical level. See fan and thermal alerts and PSU redundancy lost.
Checklist
- Room temperature normally well below 35°C, with alarms on the air-conditioning.
- All equipment front-to-back; required clearances kept; rack doors ventilated.
- Blanking panels in every empty U; drive and PCIe blanks in place.
- PSU 1 and PSU 2 on separate circuits or PDUs; PSU types not mixed.
- UPS sized with Dell EIPT or HPE Power Advisor plus other devices; shutdown tested.
- Temperature and power alerts configured in iDRAC or iLO.
- For a new room or a larger build, see data centre design and build in KSA.
Skyline can supply, configure and maintain Dell, HPE and Huawei servers under an annual maintenance contract (AMC). Contact us.
Comments
0 total · 0 threads