Other industries · Critical facilities
Data Centre FM: Chillers, UPS and Uptime
By the humblx team6 min read
In short
Data centre facility management is about proving that every redundant cooling and power path will work when called on, and maintaining equipment without dropping below contracted redundancy. It relies on disciplined chiller, UPS, battery and DG maintenance, load testing, written procedures for every switching operation, and tracked responses to every critical alarm.
- Redundancy only protects you if every path is genuinely available, so maintenance must verify it, not assume it.
- Load-test DG sets and check battery health regularly; a weekly idle start proves little.
- Use written procedures and change control for any work that reduces redundancy.
- Every critical BMS alarm should create a tracked work order with an owner and response target.
In a data centre, a facility fault does not arrive as a complaint. It arrives as an incident report with the client's legal team copied in. A failed fan on a cooling unit, a UPS battery string that has silently lost capacity, or a DG set that cranks but never takes load during a grid dip in Navi Mumbai can cost more in one afternoon than the annual maintenance contract for the whole site. That is why data centre facility management is less about fixing things fast and more about proving, every single day, that the backup to the backup will actually work.
What makes data centre facility management different?
Most commercial buildings tolerate a short failure. A data centre is designed on the assumption that failures will happen and must not matter. Cooling and power are built with redundancy, described as N (just enough capacity), N+1 (one spare unit) or 2N (a full duplicate path). The facility team's real job is to keep every one of those redundant paths genuinely available, and to carry out maintenance without ever dropping below the redundancy promised to customers.
- Availability is contractual. Customer SLAs are written in terms of uptime, and breaches carry penalties.
- Maintenance is change. Every switching operation is planned, approved and executed step by step.
- Documentation is evidence. Clients and auditors expect to see test results, service records and incident reports.
- Energy is a commercial metric. Power usage effectiveness (PUE) affects operating margin and customer perception.
Cooling: chillers, CRAH units and airflow
Indian sites face high ambient temperatures in cities such as Chennai and Hyderabad, coastal humidity in Mumbai and dust almost everywhere. Cooling plants work hard for much of the year, and small losses in efficiency or capacity add up quickly.
- Trend chiller approach temperatures and clean condenser tubes or coils before fouling erodes capacity.
- Keep cooling tower or dry cooler water treatment, drift eliminators and fill media in good condition.
- Change CRAH or CRAC filters on pressure drop and inspect fans, belts and drives on schedule.
- Check hot-aisle or cold-aisle containment: blanking panels in empty rack spaces, sealed cable openings, doors that close.
- Monitor rack inlet temperatures, not just room averages, to catch local hot spots.
- Test leak detection under raised floors and around cooling pipework.
The HVAC preventive maintenance checklist covers the routine tasks for chillers, cooling towers and air handling units, and adapts readily to precision cooling.
Power: UPS, batteries, DG sets and switchgear
The power chain runs from utility incomers and transformers through LT panels and transfer switches to UPS systems and rack distribution. Each link has its own failure modes and its own maintenance rhythm. The frequencies below are illustrative; follow OEM guidance and your site's design.
| Asset | Typical failure mode | Maintenance focus | Illustrative frequency |
|---|---|---|---|
| UPS modules | Capacitor and fan ageing, firmware or control faults | Alarm-log review, visual checks, OEM service, transfer tests | Monthly checks, annual OEM service |
| Battery strings | Cell degradation, loose terminals, overheating | Impedance or conductance testing, terminal torque, room temperature | Monthly to quarterly |
| DG sets | Failure to start or take load, fuel contamination | Daily checks, regular test runs, periodic load-bank tests | Daily, weekly and half-yearly |
| Transfer switches and switchgear | Contact wear, relay mis-settings, hot joints | Thermography, breaker servicing, transfer testing | Quarterly thermography, annual servicing |
| Diesel storage | Water and microbial growth in stored fuel | Tank sampling, fuel polishing, day-tank level checks | Monthly to quarterly |
Electrical panels deserve the same rigour. Loose connections and overheated busbars are a common cause of avoidable outages, and periodic thermography catches them early. The electrical panel maintenance checklist is a useful starting point.
How do you maintain equipment without dropping redundancy?
Take an illustrative colocation site in Hyderabad with 2N UPS. The B-side UPS is due for its annual service, so for the duration of the work every rack will depend on the A side alone. That window is the riskiest moment of the year for those customers, and it needs to be treated that way.
- Write a method of procedure listing each switching step, who performs it, who verifies it and how to roll back.
- Get change approval and notify customers within the agreed notice period.
- Before starting, confirm the A path is healthy: UPS alarms clear, batteries recently tested, DG ready.
- Freeze all other work on the A path during the window.
- Execute step by step with a second person verifying each action.
- After the work, confirm both paths are restored, redundancy is back and alarms are clear, then close the change.
Monitoring, alarms and the human response
Most data centres have a BMS or DCIM tool generating alarms. The weak point is rarely detection; it is response. Alarm floods train people to ignore notifications, and a critical alert acknowledged on a screen but never assigned can sit until it becomes an outage. Every critical alarm should create a tracked work order with a named owner and a response target, and the time from alarm to resolution should be measured as MTTR. The maintenance downtime calculator helps convert that history into uptime figures, and our article on facility management SLAs and KPIs explains how to set targets.
Vendors, spares and shift handovers
Most Indian data centres depend on several OEM and AMC vendors: one for chillers, another for UPS, a third for DG sets, a fourth for fire suppression. Each has its own report format and response commitment, and gaps open up at the boundaries between them. Keep a single register of contracts with response times and escalation contacts, hold critical spares such as UPS fans, contactors and DG filters on site, and run a structured shift handover that lists open alarms, active changes and equipment out of service. A handover done verbally at 7 am is where many incidents quietly begin.
Energy efficiency without risking uptime
Cooling is usually the largest non-IT load, so it is where PUE improves most. The safe levers are well known: tighten containment, remove bypass airflow, raise supply air setpoints gradually within the ranges in ASHRAE's thermal guidelines and your equipment vendors' limits, fit variable speed drives on fans and pumps, and use economiser modes when outdoor conditions allow. Change one thing at a time, watch rack inlet temperatures, and keep a rollback plan.
Where a digital FM platform fits
A BMS tells you what the equipment is doing. It does not schedule the quarterly thermography, record who torqued the battery terminals, chase the DG vendor for the load-bank report or prove to a client auditor that every PPM task was done. That human workflow is where humblx Facility Maintenance fits, with preventive schedules, checklists with photo proof, vendor management and a full service history per asset. Energy and utility management is coming next.
Your data centre FM action list
- Confirm every redundant path is truly available today, not just on the drawing.
- Schedule load tests for DG sets and health tests for every battery string.
- Require a written procedure and second-person check for any work that reduces redundancy.
- Route every critical alarm into a tracked work order with an owner.
- Review containment and setpoints once a quarter for safe energy gains.
Frequently asked questions
What does data centre facility management include?
It covers the physical infrastructure that keeps IT equipment running: cooling plant and precision air units, UPS systems and batteries, DG sets and fuel, switchgear and distribution, fire detection and suppression, leak detection and physical security systems. The team maintains these assets, manages vendors, controls changes, responds to alarms and documents everything for clients and auditors.
How often should data centre UPS batteries be tested?
Most sites inspect batteries monthly and carry out impedance or conductance testing monthly or quarterly, with fuller capacity testing less often, guided by the battery manufacturer and the site's risk policy. Trending results over time matters more than any single reading, because gradual degradation is the warning sign that a string may not support its rated runtime.
What is concurrent maintenance in a data centre?
Concurrent maintenance means a component can be taken out of service for maintenance without interrupting the IT load, because another path or unit carries it. Achieving it in practice needs the design to allow it, a written procedure for each operation, confirmation that the remaining path is healthy, and a freeze on other work during the window.
How can a data centre reduce PUE safely?
Focus on cooling first. Seal containment and blanking gaps, eliminate bypass air, raise supply air temperatures gradually within equipment and ASHRAE guideline limits, add variable speed drives to fans and pumps, and use free cooling when outdoor conditions allow. Make one change at a time, monitor rack inlet temperatures closely, and keep a tested rollback plan.