[nSLUG] Power Loss and UPSs

Dop Ganger nslug at fop.ns.ca
Mon Nov 5 09:08:04 AST 2007


On Sun, 4 Nov 2007, Jim Haliburton wrote:

> I am impressed that someone on the list has UPSs that will hold 50+
> servers up for 2 hours.  That is a lot of battery power.

I have a client who has batteries to power a rack full of machines - the 
two battery containers are about 3/4 the height of a 42U rack. Apparently 
he took some flack for it, right up until Hurricane Juan hit and he was 
able to run the whole operation off the batteries for 2 days until power 
came back on, and he still had power to run for another 3 days or so.

> Can virtualization allow you to reduce the server count in times of power
> instability?

A better bet might be laptops. They're significantly lower power than 
almost any server, plus they have a built-in UPS. I've run our web hosting 
environment off laptops before now when the whole building had a scheduled 
power outage and we needed to make the UPS batteries last as long as 
possible as there was a risk the outage might last multiple days (sitting 
in the server room with nothing but some bike light leds listening to the 
UPSen beep wasn't a huge amount of fun).

> With the low price of relatively good capacity, I have found it seldom 
> useful to replace batteries.  The older UPSs are likely linear power 
> supplies with large transformers etc.  New ones are far more efficient 
> and give longer run times with the same amp hour capacity batteries.

I looked into replacing the batteries on one of my older UPSen and it was 
only around $20 cheaper than a new UPS. As you say... It was a better deal 
just to buy a new UPS.

> One person mentioned putting car or truck batteries in their UPSs to
> extend their run time.  That may not be a good solution.  The batteries
> in most UPSs are designed with slightly different chemistry and a
> different charging plan than the batteries used in automotive use.  I
> have found when trying this, that the automotive batteries die in a short
> time from the charging that UPSs do.  In most cases less than a year of
> life.  It has all to do with the charging.  I was told that the UPS
> battery uses a "float" charge.  I do not know what the difference is
> between that and what a standard charger for automotive batteries does.

It's not the charging so much as the deep discharge that kills car 
batteries as the plates suphate. There's also issues of the car battery 
leaking hydrogen. Amusingly, Dan dealt with the potential issue of 
acid leakage by putting the battery on a tray with a layer of sodium 
bicarb.

The batteries don't have a huge lifetime, although I think a year is a bit 
short - I'd look at more like 2 years. This isn't far off of the typical 
lifetime for a cheapo UPS's lead acid battery anyway.

For the real enthusiasts, he built up a complete UPS from an inverter, 
bench power supply and car battery, with more ruminating on risks on 
runtime. http://www.dansdata.com/diyups.htm for the gory details.

> I look forward to hearing the experinces of others over this week-end.

The worst that happened was I lost the siding on my house, thanks to the 
numpties at Greater Homes deciding it was sufficient to nail the slats 
into the pressboard rather than, say, actual wooden joists. I was mildly 
surprised when my insurer (RBC) said they'd send round a contractor that 
afternoon to take a look, and they're going to fix it today. All the 
computing equipment came through fine.

Cheers... Dop.



More information about the nSLUG mailing list