On Thu, 7 May 2009, Tom Limoncelli wrote:

> Here's the most concise summary of what I know about power:
>
> 1.  90% of all power outages are less than 10 seconds.  A UPS can
> handle that fine.
> 2.  9% of all power outages are between 10 seconds and a few hours.  A
> UPS can handle that fine.
> 3.  1% (or less) of all power outages are multi-day outages.  UPSs
> that can handle that are way more expensive than you can afford.  Get
> a generator and use the UPS to cover the gap between when the outage
> starts and when the generator is providing clean power.
> Therefore...
> 4.  If you can't afford a generator, get a UPS that lasts lonog enough
> to shut down your machines.
> ...and...
> 5.  Because outages tend to be very small or very long, if you reach
> the 1-hour point, start powering off machines and send employees home.

what I was looking for is a little bit further.

I have people (both in-house and in opensource projects) making claims 
that they should not need to plan around the possibility of there being a 
power failue, because anyone competent will have a UPS.

my reaction to that is that I've seen entire datacenters go dark, even 
with redundant UPS. I've seen enough other people say the same thing that 
I'm confident that it's not just a case of my facilities people being 
incopentent ;-)

you can (and should) put effort into a good UPS system, and most of the 
time it will save you when there is a power problem, but it won't save you 
all the time, and if you are designing something that's supposed to be 
reliable, you can't dismiss power events with 'there will be an UPS, so we 
don't have to worry about that'

I've seen top tier brand servers with redundant power supplies shut down 
because the power supply system inside the computer flaked out and decided 
that both power inputs were bad (disabling both power supplies)

I've seen rats in the wall chew into power lines, shorting them out and 
downing the system.

I've seen UPS systems that were reviewed by the maintinance company a week 
prior fail to handle a loss of power and shut everything down

and other things.

I thought the discussion I had on this topic was here, but apparently not.

David Lang
_______________________________________________
Discuss mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to