Your UPS shows 1h30 of autonomy. You’ve built your continuity plan around this figure. On the day of a real power cut, you may have 45 minutes.
At Infologo, an IT service provider based in Geneva and Lausanne, we’ve discovered that this scenario is much more common than you might think, and that it’s due to a very precise methodological error during calibration.
This distorts the measurement without you even realizing it
UPS calibration only reflects the actual load if the UPS alone is carrying that load during the test. This is where the problem arises with dual-powered servers.
A server equipped with two redundant power supplies can receive power from two different sources. In the vast majority of installations we manage, one power supply is connected to the UPS, the other directly to the mains.
Legitimate objective: redundancy.
Problem: the server doesn’t consume both power supplies simultaneously. It chooses one or the other according to its internal logic, often the “cleanest” or the first detected.
Result: during a discharge test, the UPS only sees part of the server’s real power consumption. Its discharge curve is slowed down. It calculates autonomy on a reduced basis. The figure displayed may be double, or even more, than it would be under full load.
We have seen this configuration in around 80% of customers whose servers are managed under outsourcing contracts. This is no exception.
Why isn’t this an inverter error?
The inverter does what it’s told. It measures what it sees. If the actual load during the test is not representative of the load during a real outage, its calculations will be consistent with its measurements, not with reality.
The confusion arises from the fact that most technicians perform the calibration without disconnecting the parallel power supply. This detail goes unnoticed, as the server remains operational throughout the test. Everything seems normal. Calibration ends, and an autonomy figure is displayed, as noted in the documentation. No one questions the validity of the measurement.
This problem is not specific to any particular UPS brand or server model. It applies whenever one of the two power supplies of a dual-powered server is not on the UPS at the time of testing.
The impact on your business continuity plan
An IT continuity plan is based on known times: how long does the infrastructure run on battery power, how long does the generator take to start up, what is the acceptable transfer window. If the actual autonomy of your UPS is 30% to 50% less than you’ve measured, the window may close before the generator takes over. Or before your teams have had time to initiate a clean system shutdown.
During a recent audit at a customer site in the Morges area, the calibrations on four inverters were re-examined. None had been carried out under the correct conditions. Continuity documentation quoted durations that had never been validated in real-life conditions.
What we recommend to our customers at Infologo
For a calibration to be valid on a dual-powered server, the parallel power supply must be disconnected for the duration of the test. The server must run solely on the UPS, with the effective load it supports in production.
This is not a trivial operation: it requires a maintenance window, coordination with the teams using the systems concerned, and verification that the switchover to the UPS power supply is carried out properly before launching the test. In some cases, a partial shutdown of services is programmed to limit risk during the window.
The procedure must be documented and identical for each person involved. A calibration done “by hand” by three different technicians using three slightly different methods will produce three different figures, none of which will be reliable.
If you’ve had UPSs in place for several years, and the calibrations were carried out on servers with dual power supplies, without disconnecting the mains power supply, consider these figures to be invalid. Repeat the tests under the right conditions before your next business continuity plan review.
Are you managing an infrastructure with UPS on dual-powered servers and haven’t yet checked your calibration conditions? Contact us for an audit, and we’ll look into it together.
