Choosing the right Uninterruptible Power Supply (UPS) is critical to prevent data corruption and downtime. This guide breaks down the confusing specs like VA vs. Watts and Pure vs. Simulated Sine Wave to help you protect your servers effectively.

You Bought a UPS. Your Server Still Died. Let’s Talk.

I still remember the 3 AM PagerDuty alert. A micro-outage, maybe two seconds long, caused by a squirrel deciding a transformer looked like a tasty snack. The rest of the rack was fine, but prod-db-01, our primary PostgreSQL server, went down hard. We had a UPS on it, a cheap one the previous admin had bought. The problem? The server’s high-efficiency power supply saw the UPS’s messy “simulated” sine wave output, panicked, and shut down instantly. What followed was four hours of restoring from backups and explaining to management why our “uninterruptible” power was, in fact, very interruptible. Don’t be that person. Let’s get this right.

Why Your “Uninterruptible” Power Supply… Interrupted

The core of the problem isn’t just about having a battery; it’s about the quality of power that battery provides and how your server’s Power Supply Unit (PSU) reacts to it. Most modern servers use Active Power Factor Correction (PFC) PSUs, which are incredibly efficient but also incredibly picky about the electricity they consume.

• VA vs. Watts: This is the classic trap. VA (Volt-Amps) is “Apparent Power,” while Watts is “Real Power.” Think of it like a glass of beer: VA is the total volume of foam and beer, but Watts is just the actual beer you want to drink. Your server’s PSU is rated in Watts. A cheap UPS might have a high VA rating (e.g., 1500VA) but a low Wattage rating (e.g., 900W). If your 1000W server is plugged into it, it’s going to fail under load.
• Simulated vs. Pure Sine Wave: This is the silent killer. Wall power is a smooth, clean “Pure Sine Wave.” High-end UPS units replicate this perfectly. Cheaper, “Simulated Sine Wave” units create a chunky, stepped approximation of that wave. An Active PFC PSU can interpret these blocky steps as a power fault and will shut down to protect itself—completely defeating the purpose of the UPS.

Solution 1: The “It’s Just Plex” Quick Fix

Look, sometimes “good enough” is good enough. If you’re just trying to keep your home router, a simple NAS, or a non-critical server from rebooting during a flicker, a modern Simulated Sine Wave UPS from a reputable brand (APC, CyberPower, Tripp Lite) will probably work. The PSUs in consumer-grade gear are often more tolerant.

How to do it:

1. Find the max wattage of the devices you want to protect. It’s usually printed on the power brick or PSU sticker.
2. Add them all up. Let’s say your NAS is 150W and your router is 20W, for a total of 170W.
3. Buy a UPS with a wattage rating at least 20-25% higher than your total. So for 170W, you’d want a UPS rated for at least ~215W. Ignore the VA rating for this calculation.

Warning: I call this the “quick fix” for a reason. I would never trust this setup for a production database, a critical virtualization host, or anything that could corrupt data on a hard shutdown. You’re gambling that your PSU is tolerant enough.

Solution 2: The Professional’s Choice (Pure Sine Wave + NUT)

This is the correct way to do it for any system you care about. You invest in a Pure Sine Wave UPS and configure software to gracefully shut down your machines when the battery gets low. This protects your hardware and your data.

A Pure Sine Wave unit guarantees that your picky Active PFC power supply gets the clean power it expects, preventing an immediate shutdown. Then, you use software like Network UPS Tools (NUT) to manage the outage.

NUT runs a server on a machine connected to the UPS via USB. Other machines on the network (clients) can then monitor the UPS server’s status. When the battery hits a critical level, the server tells all the clients to perform a clean shutdown.

Here’s a simplified look at a NUT client config (upsmon.conf on a client machine like proxmox-host-01):

# Monitor a UPS defined on the machine 'ups-server'
# The format is @    
# '1' means this machine requires 1 power supply from this UPS.
# 'master' means the ups-server machine is the master, this one is a 'slave'.

MONITOR myups@ups-server.lan 1 upsuser mysecretpassword slave

With this setup, when `ups-server` reports a low battery, `proxmox-host-01` will automatically receive the signal and run `shutdown -h +0`.

Solution 3: The “Zero Downtime” Nuclear Option

In a real data center or for a critical business service, we can’t afford any downtime. This is where you bring in the heavy artillery: rack-mountable, network-managed UPS systems, often with external battery modules (EBMs) for extended runtime.

Units like the APC Smart-UPS or Eaton 5P series are the industry standard. The key features here are:

• Network Management Card (NMC): This gives the UPS its own IP address. You can log into a web interface to see stats, trigger self-tests, and configure alerts without needing a dedicated server connected via USB.
• Extended Runtime: You can daisy-chain multiple battery packs to provide hours, not just minutes, of runtime.
• Power Chaining: For critical servers with redundant PSUs, you plug one PSU into a UPS on Circuit A, and the second PSU into another UPS on Circuit B. This protects you from a UPS failure or a tripped breaker on one circuit.

Darian’s Pro Tip: Your UPS setup is worthless until you’ve tested it. I mean really tested it. Unplug the UPS from the wall and watch what happens. Does your NUT client actually trigger the shutdown? Does the server power back on correctly when power is restored? Don’t wait for a real outage to find the bug in your shutdown script.

Which Path Should You Choose?

To make it simple, here’s a breakdown.

At the end of the day, match the solution to the problem. Protecting your kid’s Minecraft server doesn’t require a $2000 rackmount unit. But protecting your company’s primary database on a $100 budget special is just asking for a 3 AM data recovery fire drill. Spend the extra money on a Pure Sine Wave unit. You’ll thank me later.

Darian Vance

Lead Cloud Architect & DevOps Strategist

With over 12 years in system architecture and automation, Darian specializes in simplifying complex cloud infrastructures. An advocate for open-source solutions, he founded TechResolve to provide engineers with actionable, battle-tested troubleshooting guides and robust software alternatives.

🤖 Frequently Asked Questions