power supply - Why is UPS technology unreliable?

Following the UPS issue at SO today I recall my experiences with UPS (uninteruptable power supply) systems in the work place.

Typically we would arrive at work in the morning and find all systems out. There was a power outage and the UPS failed to work correctly. In fact I can only remember one time when a UPS actually worked. That's a pretty poor track record - though of course my own experiences may be just unfortunate and UPS systems in reality are as reliable as having two slices of bread at opposite sides of a cheese sandwich.

EDIT The original question may invite opinion based answers. Apologies - my question is rephrased now as:

What factors can cause UPS systems to fail? What are the limitations of UPS?

To directly answer your question, not if they are good quality, used properly and within spec.

I have not had bad experiences with UPSes when they are used properly. Much of this depends on your specific situation. While IMHO the way you have worded your question invites extended discussion rather than precise answers and might be considered a poor fit for Superuser's Q&A format, there are some definite non-argueable points that one should consider when relying on UPSes.

• Are you using name brand UPSes?

• Are they new? Buying used UPSes is generally a bad idea.

• Is everything connected under the wattage limit for the UPS?

• Are you in an area with poor quality power or lots of storms? Those can shorten the life of any UPS.

• Desktop UPSes are generally meant to give a system 15 or 30 minutes of uptime to allow for clean shutdown. Solutions that keep an entire datacenter running are typically diesel run generators and things like that. Are your computers configured to shutdown when the UPS is low on battery?

• Batteries and UPSes wear out and must be refreshed periodically. So if you have a system on a 5 year old UPS it might not be as reliable as a new one.

• If you are depending on UPSes as part of a business continuity plan, you really should buy a few, connect them, configure them properly, and then test them to be sure they will work according to your expectation. Rolling out something like this without a testing phase to confirm may lead to a bad time.

I suspect it rather depends on a number of factors including:

• What UPS? Is it perhaps under-specified for the task? Or were corners cut in procuring it?
• How reliable is the incoming power source? It it perhaps very poor quality (brown outs?) causing the UPS to constantly kick in for a short period over night - I see that happen on my own home-office UPS sometimes. Could that be draining the battery?
• What kind of load do you have on the UPS?

I've found good quality comms room and data centre UPS's to be very reliable indeed over the years. The most common failure being to forget to fill up the diesel fuel after an outage!!

Home and small office UPS's are also highly reliable, indeed I was able to work from home for a week or so after the Sheffield "Great Flood" a few years ago with the power going on and off like mad. However, I've found that the batteries can fail. I now only have one instead of two working (purchased at the same time) because one of the batteries failed after a couple of years.

No, of course they need to have very good reliability. The problem is the old curse of the redundancy: to make things redundant, you need a system which integrates the childs. For example, in a hw raid array, you need a raid controller. If this controller dies, your whole array dies, although it was redundant.

Or if you try to make redundant servers, you need something which finds out which server is unavailable and redirects traffic to its clones. This thing can also die.

There are completely redundant systems possible, but they are normally too complex, and most sysadmin simply won't do that. Most of them simply don't even understand the concept. In most cases, they are using not redundant, but very reliable systems to connect and manage the redundant cluster of not so reliable sub-systems.

For example, there is a very costly raid controller, using a redundant disk array from the far east.

It is nearly always the case, but this is which won't work for UPSes. Why?

1. UPSes are working with current. If the primary power source is lost, the secondary need to be powered on in 1/100 - 1/10000 seconds, or the outer voltage of the internal transformators of the servers will significantly drop.
2. UPSes can't be tested. You can't plug out the power to find out if an UPS is okay, because your server will be shut down if it isn't.

For real ups redundancy you needed to use a redundant server cluster using servers with redundant power sources.

In most case, in big server farms, it is not done. Yes, google will also die if the actual server node dies which you are actually using. But they will find the bad nodes very fast, and eliminate that from their system.