I'm looking for recommendations on UPS features or models that can reliably keep servers online during storms. Several customer systems have shut off instantly when the power flickers instead of switching to battery. One server is connected to two separate 1000VA tower UPS units—one APC and one CyberPower—and both have failed during outages more often than they have successfully carried the load. They do work when manually unplugged from the wall.
Could this be caused by surges, poor power quality, aging batteries, insufficient capacity, or the UPS design? The units may be basic standby or line-interactive models, and I'm wondering whether an online/double-conversion UPS is necessary for critical servers. Advice on sizing, battery maintenance, and dependable models would also be helpful.
5 Answers
Line-interactive units are a useful middle ground and many work well for ordinary computers and network gear, but they still rely on a transfer when power leaves the acceptable range. For genuinely sensitive or mission-critical servers, double-conversion is the safer choice. Whatever design you buy, size it based on measured load and required runtime, maintain the batteries, and budget for eventual UPS replacement rather than treating the unit as permanent infrastructure.
Before replacing everything, check the batteries and the actual UPS logs. A battery can pass the unit’s self-test yet collapse under load, especially if it is several years old. UPS batteries commonly need replacement around every three years, depending on temperature and usage. Also verify that the equipment is plugged into the battery-backed outlets rather than surge-only outlets, and keep the normal load comfortably below the UPS rating—around 60% is a reasonable target.
Good units should record why they switched off—overload, battery failure, output fault, excessive input voltage, or a transfer problem. Models with network management and event logging are valuable for diagnosing intermittent storm issues. It is also worth having an electrician check the building’s grounding and install appropriate upstream surge protection; a UPS is not a substitute for correcting serious electrical problems.
A 1000VA unit may be undersized for a server, particularly if there are multiple power supplies or other equipment involved. Measure the real wattage draw instead of relying only on the power-supply label, and make sure each UPS can handle the load during the transfer event. If two UPS units are feeding separate server power supplies, confirm that the server and UPS arrangement is supported and that neither unit can become overloaded when the input power changes.
For a critical server, the key term to look for is online or double-conversion UPS. It continuously converts incoming AC to DC and then back to clean AC, so the load does not have to wait for a relay to switch during a disturbance. That makes it much less sensitive to brief dropouts, surges, and unstable utility power. Eaton 9PX, Eaton 5PX, and Vertiv/Liebert GXT5 are examples people report good results with, although the right model depends on the load and runtime you need.
That makes sense. The existing units may be basic standby models, so I’ll compare the cost of replacing them with double-conversion equipment rather than assuming another inexpensive tower will solve it.
We had servers reboot during brief flickers even though the UPS reported a healthy battery. Replacing the batteries fixed it, so battery age is definitely worth checking.