Editor's Note: ScreenJet's founder Alan Yeo wraps up his investigation of UPS units, having had a pair fail and then take two HP 3000s offline recently. Here he explains what sort of UPS to buy to avoid a failure that knocked solid 3000s offline, by way of dirty transfers during the all-important Transfer Time (TT) window.
By Alan Yeo
Last in a series
First off, the answer to the problem: Double Conversion UPS units are what you want. They are more expensive than Line Interactive ones, but it is claimed they are cheaper in the long run, due to increased battery life. I’ll let you know in a few years. The HP 3000
Whilst a Line Interactive UPS claims that attached equipment shouldn't be at risk during the TT window, as far as I can read it can be before it is disconnected from the mains. APC for example have a compensation scheme which wouldn't be required if this wasn't possible. Note: It's interesting that APC only offer this protection policy on 120V products, for those of us using 220/240V supplies the risk is obviously deemed to be too great to cover. The fine print:
If your electronic equipment is damaged by power line transients on an AC power line (120 volt) while directly and properly connected to a standard APC 120 volt product covered by the Equipment Protection Policy (EPP), you can file a claim with APC for compensation of your damages. Coverage of damages is determined by the limits of the EPP.
TT seems to be related to the Sensitivity Level: High, Medium or Low. And the Sensitivity Levels can be altered by how you configure the UPS, for example on many UPS's you can adjust at what upper and lower input voltages it should transfer to/from battery. On 120V UPS's this range is typically 127-136 at the upper end, and 97-106 at the lower end. In High Sensitivity mode the TT is something like 2 milliseconds and if set Low around 10 milliseconds. Switching to/from battery frequently is bad for battery life, as is protracted running from the battery on a Line Interactive UPS. So the compromise is between High Sensitivity with possibly frequent but low TT, and Low Sensitivity with less frequent but longer TT.
Theoretically during the TT there is no power going to the connected equipment, so long TT's may be a problem for some equipment, also if transfers are frequent equipment may see a pulsed power supply.
If Line Interactive UPS technology can fail, what should you use? On-Line or Double Conversion technology seems to be the answer. These can be hard to spot as the word “Online /On-Line” is used to describe a mode of virtually any UPS: i.e. a Line Interactive UPS is described as being in online mode when it is feeding mains direct to the attached equipment. So it's probably best to use the term Double Conversion, as this is less misleading. In a Double Conversion UPS the equipment is always fed from the inverter, which has tandem input supplies, one from the battery and one from a mains fed rectifier. This means that the equipment never sees “Mains” power, and there is never any TT as the supply from the inverter is continuous regardless of which power source it is using.
Double Conversion UPS's are more expensive to buy than Line Interactive ones, but it is claimed they are cheaper in the long run due to increased battery life. I'll let you know in a few years.
In just checking a few facts I have just discovered that Wikipedia has a great page that clearly covers the different types of UPS technology. So much so that I wish I had found it before, and also hadn't bothered trying to write this explanation. So if you want more info, or are totally confused by my description, try: http://en.wikipedia.org/wiki/Uninterruptible_power_supply . There is also a great white paper, The Seven Types of Power Problems in PDF format on the Web.
Wow, are we an HP customer again?
Surprise, Surprise. Apart from the odd printer cartridge, we haven't bought anything HP since 2001, and I couldn't see much likelihood that we would. But we have just bought a new HP UPS! It appears that as a result of the “Invent” phase (which I understand has now been terminated) HP are one of the leaders in advanced UPS technology, and they do a range of Double Conversion units. We were fortunate and picked up a really nice HP unit, plus a huge auxiliary battery pack (we now have an estimated 2 hours full load uptime) for less than the cost of a smaller new APC Line Interactive. Okay, they were customer returns due to damaged packaging, but were brand new and unopened.
The sad part for HP is why we got it so cheap! I had to ask the UPS reseller why the HP one was a lot cheaper than other makes of similar Double Conversion UPS's he had for sale, especially as the new list price was as high or higher than the others. His answer surprised me.
“They are difficult to sell, because nobody recognises HP as a UPS supplier.” He told me that customers who ran data centers with HP Servers would buy new HP UPS's, but that in his experience nobody else did. So if they got returns or cancelled orders, they found them very hard to shift unless they discounted them heavily. He did say that in his opinion that it really was nicely-made equipment, and that we were getting a great deal (well he is a salesman!).
It's an ill wind that blows no good
Well the wind may have caused this, but its certainly blown away a few misconceptions we had about how protected we were and how good our backup recovery strategy was. I think we had OK strategies for either total loss, or losing a single Server to a specific problem. But I don't think we had anticipated multiple (but not total) failures at the same time, or an HP 3000 outage that was caused by multiple problems that could only be discovered in a serial manner. Our total recovery time was days, not hours! The upside is that we are now better protected, have less kit running and plans for even less (feet on the ground, head in the cloud) and are now working on a recovery/disaster plan that encompasses what we have learnt.
Hopefully this saga may be a warning to others to pull out the manual for your UPS and see if it really meets your needs and expectations.
Oh by the way, on the Saturday after the outage I had a visit from an engineer from the local power distribution company, to check out our voltage that had been running earlier that day at about 264V (standard here in the UK is 240). He said our problem spikes were probably caused by a commercial neighbour with a badly-configured backup generator setup. Or it could have been the power company themselves using one to fill in a hole in the grid due to downed lines. But that we would never be able to prove it! Anyway to finish, here’s a couple of nice quotes I found at the APC site:
How large can a surge be?
Electrical industry standards indicate that electrical power surges inside a building can reach levels up to 6,000 volts and 3,000 amperes, which could deliver up to 90 joules of energy.
How often do electrical surges occur?
Very large surges could occur a few times a year in medium exposure areas or as often as 40 times a year in high exposure areas. All of which may be storm induced. Beyond storm induced electrical disturbances, normal equipment operation can also produce surges, some over 1,000 volts. These surges may occur several times a day.
About that Dual Conversion claim of increased battery life, I’ll let you know in a few years. HP 3000s are certain to still be running by then!
Alan Yeo is a developer and entrepreneur at ScreenJet, which delivers the TransAction any-platform replacement for Transact, as well as ScreenJet software, plus interface modernization services for HP 3000s which rely on VPlus today.