Data centre design reference

Generic reference designs for power, cooling, controls and operations

Operations and maintenance

Operating and maintaining a data centre: routines, spares, procedures and records.

On this page
  1. One register from design to operations
  2. A maintenance programme built from failure modes
  3. Design for the people who run it
  4. Commissioning and handover
  5. Sources

One register from design to operations

Most facilities are handed over with drawings, a 3D model, a control-system point list and a maintenance system that each name the same pump differently. Operations then spends its first year reconciling them. The reference approach gives every asset one key at design and carries it everywhere.

The register-keyed digital twin: one asset register joins the 3D model, live data, maintenance and documents Every physical asset gets one register key. The 3D model, the control system's tags, the maintenance system's work orders, the documents and the label on the equipment all carry that key, so any of them leads to the others. Asset register one key per asset 3D model IFC, element ids Documents drawings, manuals, tests EPMS and BMS points, alarms, trends CM CM live condition data Maintenance system tasks, history, spares Equipment label QR code with the key 1. The key is assigned once, at design, and carried through construction, commissioning and handover, so operations start with a complete register instead of rebuilding one.
The register-keyed digital twin: one asset register joins the 3D model, live data, maintenance and documents

Design position. The register key is a design deliverable, like the single-line diagram. It goes on the drawing, the equipment label, the control-system tag and the maintenance record, and the handover is checked against it.

A maintenance programme built from failure modes

The programme starts from how each class of equipment fails, how that failure shows itself beforehand, and how long the warning lasts. Tasks and inspection intervals follow from that, not from a generic calendar. The method follows IEC 60812 for failure modes, ISO 14224 for the equipment taxonomy, and ISO 17359 and ISO 13374 for condition monitoring.

EquipmentWhat to watchHow
Circulation pumps on drivesSeal leaks, bearing wear, cavitation, standby that will not startVibration against ISO 20816 zones (formerly ISO 10816), thermography, scheduled duty and standby swaps
Plate heat exchangersFouling, gasket leaks, cross-contaminationApproach temperature and pressure-drop trends, samples from both circuits
Strainers and raw-water screensBlockingDifferential-pressure trend, back-flush count
CDUs and the technology loopFilter loading, coolant chemistry, leaksCDU flow and pump alarms, laboratory analysis of the coolant, leak detection
Water treatmentCorrosion, microbial growthLaboratory analysis (pH, conductivity, inhibitor, metals, microbiology), corrosion coupons
UPS and batteriesWeak cells, hot connections, transfer faultsOhmic trending, capacity test on a load bank, thermography, transfer tests
Busway and tap-off boxesLoose or hot jointsThermography (ISO 18434-1), contact-resistance measurement
Variable-speed drivesFan and capacitor ageing, hot connectionsFault log, heat-sink temperature trend, thermography
Valves, check valves, air ventsHidden failures: a valve that will not close, a check valve that passesPlanned exercise and closure tests, reverse-rotation check at pump changeover

The maintenance standards named above (IEC 60812, ISO 14224, ISO 17359, ISO 13374, ISO 20816 and ISO 18434-1) are named for orientation. The research behind this site did not review them.

NFPA 70B, the electrical maintenance document many teams follow, became a standard in its 2023 edition, the edition the sources here describe; check which edition applies to your project. Trade sources point out that it becomes enforceable when an authority or a contract adopts it.12 On the management side, the part of ISO/IEC 22237 that covers operational information was, at the time of writing, still a technical specification rather than a full international standard.3

Design for the people who run it

Uptime Institute's 2025 outage analysis found that nearly 40 % of organisations had suffered a major outage caused by human error over the previous three years, and that the share of human-error outages caused by failure to follow procedures rose by ten percentage points.4 Uptime says the rise may be a consequence of rapid growth and staff shortages, without putting a number on it.4

Design can lower that risk: labels that match the procedures, isolation points that are easy to find and lock, procedures linked to the asset in the 3D model, and alarms that name the first action.

Commissioning and handover

The five commissioning levels run from factory acceptance through installation checks and start-up to functional performance tests and the integrated systems test, which runs the plant at design load on load banks and drops the utility supply.5 The reference handover adds one more check: every asset in the register is found on site, labelled, in the control system and in the maintenance system, before operations sign.

Sources

  1. NFPA 70B changing from guideline to standard. HFM Magazine, 2022-12-16. Read 2026-10-18.https://www.hfmmagazine.com/articles/4635-nfpa-70b-changes-from-a-guideline-to-a-standard
  2. NFPA 70B Compliance: What Changed and When It Actually Applies. Reliable Media (Reliamag). Read 2026-10-18.https://reliamag.com/guides/nfpa-70b-compliance/
  3. ISO/IEC 22237 & EN 50600: Data Centre Design Standard. GreenCalculus, 2026-06-19. Read 2026-10-18.https://www.greencalculus.com/standards/iso-iec-22237-en-50600-data-centre-design/
  4. Uptime Announces Annual Outage Analysis Report 2025. Uptime Institute, 2025-05-06. Read 2026-10-18.https://uptimeinstitute.com/about-ui/press-releases/uptime-announces-annual-outage-analysis-report-2025
  5. Data center commissioning levels and process field guide. Anvilfield, 2026-06-27. Read 2026-10-18.https://anvilfield.com/field-guides/datacenter/data-center-commissioning-levels-process/