[BBLISA] UPS relative merits

[Daniel Feenberg]

On Tue, 30 Aug 2011, Mark Lamourine wrote:

> 
> 
> On Tue, Aug 30, 2011 at 11:14 AM, Daniel Feenberg <feenberg at nber.org> wrote:
>
>       We need a UPS to prevent damage to our disk drives when power
>       spikes.
> 
> 
> What kind of system are the "disk drives" in?   Are you talking about RAID
> trays or internal disks for desktops/servers?

Internal disks in tower cases.

> 
> What is the cost (materials and labor, downtime) of a failure. It doesn't
> take much to cost more than $300.
>  

That is true if the extra $300 does something to reduce failure. There 
isn't anything in the APC literature that claims that. In the literature I 
only see the LCD, the USB port and the reference to step functions. I 
know the LCD and USB port are not important to us, and I was wondering if 
the step function power was really a problem for a few minutes per year. I 
have heard it is bad on a continuous basis.

>       We
>       have been getting APC Smart UPS for ~$500, but I have noticed
>       that the APC
>       BackUPS is about $200. If we don't care about the USB interface,
>       or the
>       LCD monitor, or the "hang time", is there a reason to prefer the
>       Smart
>       UPS? I notice in the spces that it delivers "sinewave power"
>       rather than
>       "step function approximation". Is that important for the 10
>       minutes/year
>       the system will be providing power?
> 
> 
> If I understand correctly the main difference in the power output between
> BackUPS and Smart UPS is that the SmartUPS output always comes from the
> battery through an inverter.  The power input passes through a rectifier
> which maintains a slight over-voltage which keeps the battery charged. The
> switching is on the input side.  When power fails, the charging circuit
> drops it's input, but there's no other affect on the output.
>
> The BackUPS has a power switching relay and an inverter but no rectifier.
>  Power feeds directly through from the input to the output.  A trigger
> circuit detects errors in the power input and switches from line power to
> inverted battery power.  There is a short (much less than one cycle) noise
> on the output.  Most computer power supplies can handle that kind of line
> noise.  The output circuit most certainly also has some noise conditioning
> circuit both to clean up the input line power and to smooth the transition
> to battery supplied AC.

That is what I thought too, but APC describes both systems as 
"line-interactive". If the description is the same, can the systems be so 
different? The author of the APC brochure may assume that the user doesn't 
care, but surely the user does care, or he wouldn't be spending an 
additional $300.

> 
> if you're supporting disk trays and have other means of powering them down
> on a long duration (> 10 secish) line failure and then my tendency would be
> to say stick with the SmartUPS.  The remaining danger is a line spike making
> it through to the output before the trigger can fire and isolate the output.
>  WIth the SmartUPS that's no an issue, because the output never switches,
> and input line noise never (??) reaches the output.

That has been our reason to use the SmartUPS.

> 
> If the systems are critical, or if failures and the resulting downtime and
> maintenance costs are common, I'd stick with the SmartUPS.
>

Downtime during power failures is not a problem for us. Damaged data is a 
great nuisance, since resilvering or restoring a drive now takes hours and 
hours, if not days.

Thanks
Dan Feenberg
NBER