The power goes out. The UPS beeps reassuringly — and four minutes later the server room goes dark, because the unit was sized years ago for half of today's equipment and its batteries have quietly aged. Or the opposite: an oversized UPS keeps a printer, a coffee machine and three monitors alive, while the switch that actually matters is plugged straight into the wall.
Sizing a UPS (uninterruptible power supply) is not difficult, but it has to start from the right questions. Here is a practical method for small server rooms and network cabinets.
1. Decide what the UPS is for
Runtime requirements differ enormously depending on the job, so start by agreeing which of these it is:
- Bridge to a generator: the UPS only has to cover the minutes until the generator starts and stabilizes.
- Clean shutdown: the UPS keeps systems running long enough to shut down servers and storage in the correct order, without data corruption.
- Ride-through: critical services keep running through typical outages. Long autonomy on batteries alone quickly becomes large and expensive; beyond a certain point, a generator is the more sensible answer.
2. Decide what really needs backup power
Not everything with a plug deserves battery power. Sort the equipment into three groups:
- Must stay up: firewall and internet router, core and PoE switches (which also power IP phones, Wi-Fi access points and cameras), servers and storage, the CCTV recorder and access control controllers.
- Needs only a clean shutdown: workstations with unsaved work, some lab or production PCs.
- Should not be on the UPS at all: laser printers and copiers (high inrush current), heaters, kettles, coffee machines and air conditioning. Cooling is a real issue in long outages, but it belongs on a generator, not on UPS batteries.
3. Measure the load in watts, not VA
UPS units are rated in both volt-amperes (VA) and watts (W). The watt rating is usually equal to or lower than the VA figure, and it is the one to size against. Measure the real draw of the equipment — through metered PDUs, the existing UPS display or a clamp meter — instead of adding up power supply nameplates, which overstate typical consumption. Remember PoE: a switch feeding thirty cameras draws far more than the same switch at idle.
Then add headroom. A UPS running close to its full rating leaves no room for growth and shortens runtime. A common rule of thumb is to keep the design load at no more than about 70–80% of the UPS watt rating.
4. Estimate the runtime
A simple first estimate: runtime (hours) ≈ usable battery energy (Wh) × inverter efficiency ÷ load (W). Battery energy is the battery voltage multiplied by its amp-hour capacity, summed across the whole battery string.
For example, a hypothetical 1,000 Wh battery bank with 90% inverter efficiency feeding a 600 W load gives 1,000 × 0.9 ÷ 600 ≈ 1.5 hours in theory. Halve the load and the theoretical runtime doubles — which is why removing non-essential equipment from the UPS is the cheapest runtime upgrade there is.
To put the steps together, consider a hypothetical network cabinet: a firewall, two PoE switches feeding phones, access points and cameras, a small server and a CCTV recorder, measured at about 900 W in total. With a 75% loading target, that points to a UPS rated at least 1,200 W. If the goal is a clean shutdown plus a margin for short outages, the runtime target might be fifteen to twenty minutes at 900 W — checked against the manufacturer's table, with aged batteries in mind. If the business wants phones and cameras to keep working through long outages, it is better to separate that load — a dedicated UPS or a generator for the PoE switches — than to oversize batteries for everything.
5. Choose the right topology
- Offline (standby): switches to battery when the mains fails. Inexpensive; suitable for individual workstations, not for server rooms.
- Line-interactive: regulates voltage without going to battery for every fluctuation. A common choice for network cabinets and small server rooms with reasonably stable power.
- Online (double conversion): equipment is always fed from the inverter and isolated from mains disturbances. The usual choice for servers and storage, and where power quality is poor; it runs warmer and costs more.
6. Shutdown, monitoring and maintenance
A UPS that is not connected to anything simply runs until the batteries are empty, and then everything drops at once. Connect it to the network, configure automated shutdown in the right order, and test the whole sequence — not just the battery.
- Network management card and shutdown software configured for servers, virtualization hosts and storage.
- Alerts for on-battery, low battery, overload and failed self-test, sent to people who will respond.
- Periodic self-tests and runtime calibration as the manufacturer recommends.
- A battery replacement plan based on the manufacturer's design life and test results.
- A maintenance bypass, so the UPS can be serviced without switching off the load.
- A load review every time equipment is added to the rack.
Monitoring and battery care are easy to forget once the unit is installed; they fit naturally into ongoing IT support, alongside the rest of the network infrastructure.
Not sure whether your UPS would carry today's load? Get in touch — we can help measure the load, check the runtime and decide what really needs backup power.