Cheap or free auto-shutdown setup (was: Reliability Sys V file sys)

Bill Kennedy bill at ssbn.WLK.COM
Fri Oct 5 14:41:51 AEST 1990


In article <80 at comix.UUCP> jeffl at comix.UUCP (Jeff Liebermann) writes:

I'm deleting a whole lot of this because it doesn't pertain to what I
have to say.  There are some potentially misleading statements about
"UPS" (uninterruptible power supplies) and while I'm not lashing the
author, I'm vigorously objecting to the misleading remarks.

[ ... ]
>
>Close.  I think most users have a rather odd illusion that the UPS is
>a generic fix for most AC power related problems.  The implimentations

A UPS is not only a generic fix for most AC power related problems, it's
a panacea.  With a true on-line UPS you have the utility powering a
battery charger and the battery being charged powering an inverter that
turns battery power into picture perfect AC power.  Emerson and other
manufacturers have promoted the notion that "standby" power supplies (SPS)
are UPS.  That's just not so.  An SPS passes the utility (most have some
noise filters in-line) straight through to the load until the utility
craters, then they switch on their inverter and make power from their
battery.  Everybody got the difference?  A UPS makes the AC your system
uses 100% of the time, SPS makes it only when the utility is gone.  Can
and do utility anomalies pass through SPS?  Definitely.  UPS?  Probably
not (though not impossible).

>for shutdown previously mentioned only work under ideal conditions.
>Last year, Santa Cruz survived an earthquake.  We learned a few new
>lessons about computer reliability.  My conclusion is that the UPS
>is only useful for MINOR power line disturbances.              SPS agreed!

>1.  The power died about 20 seconds after the quake started.  During
>these 15 seconds, the power was there but oscillating wildly.  Most
>UPS's did NOT activate during this time.  The high frequency trash

Is the difference becoming clear?  SPS wouldn't activate unless its
threshhold was met, UPS wouldn't activate because it doesn't need to
"activate", it's already running and AC input fluctuations aren't
interesting to a dull old battery.

>went straight thru the UPS, filters, surge protectors, switching
>power supplies, and ended up causing 2 hard disk controller cards to
>lose step and eat the data on the hard disk.

I believe it of an SPS, if it was a UPS then it was a Clary.  They are
the only ones I know who are careless enough to manufacture a UPS with
all of the genetic defects of an SPS.

Before you tune out, I'm typing this on a SunRiver workstation attached
to my 386/ix box.  This box is on an SPS, the box it connects to is on
a true on-line UPS.  SPS has its purpose, I've got two of 'em, $400 a
pop.  UPS has its purpose, I've got one of 'em, $5,000 a pop.  Am I rich?
Absolutely not!  Am I paranoid?  Absolutely!

>2.  Two customers were in the middle of an automatic backup during
[  I don't know beans about their set up, I'm sorry they lost data ]
>
>3.  Two other customers wished they didn't have UPS's.  The UPS
>kept the drive running for about 30 minutes after the start of the
>quake.  The shaking didn't stop for 5 MONTHS and we had regular
>aftershocks.  The hard disks were literally pounded into uselessness
>in the first few minutes.

There's precious little that an electrical circuit, good or bad, can
do about physical vibration.  Accordingly, I'm not sure if this is a
complaint or not, if it is a complaint, it's bogus (physical vs
electrical abuse).

>4.  Swaying power lines were a real problem.  One area had a 15Kv
>distribution line touch the local power line.  Everything turned on
>and plugged in was vaporized.  The power was also erratic for weeks.

That's one nice thing about a bona-fide UPS, it swallows all of
those anomalies into that huge multi-farad capacitor, the battery.

>The problem wasn't when the power dropped off, but when it came back
>on.  It usually appeared as a major hi-voltage spike.  MOV (metal-
>oxide-varistor) type of surge protectors couldn't handle the energy
>and either exploded or caught fire.   I lost 2 UPS's, 3 surge
>protectors, and one computer this way.

He lost two SPS's and everything else.  Let's add this up.  Let's say
that an SPS is $400, a surge protector is $20, and a computer is
$2,000.  My arithmetic says that this adds up to $2860.  It all
got (I'm sorry for that) toasted, so he gets to spend another $2860
and he's still naked!  A "real" on-line UPS, 3KVA (over 3,000 watts)
runs about $5,000.  Whoosh?  It's less than twice $2860.  It depends
on how serious you are about reliability and your data.

>These stories are only the ones related to UPS's.  My current opinion
>is that what I really want is about 1 second of power "float" from
>a large capacitor to take care of the transients.  The hard disk
>should have a good brake on the spindle.  Longer than about 1 sec.
>and the power to the computer drops IMMEDIATELY and stays off until
>the power is absolutely stable for 10 minutes.  I care less about
>trashed data, files, and such than a $$$$$ hole in my budget.  Any
>of the previously mentioned "graceful" shutdown scripts will take
>far too long to bring the system down before a quake, or the
>power company take their toll.

There's *nothing* electronic that I've seen advertised that will sense
an earthquake.  I guess you could build something up from an old Mk XIV
torpedo fuze that might start a shutdown when the tremblor started, but
a bona-fide on-line UPS would pay for itself in a hurry if you're truly
serious about a $$$$$ hole.  If you didn't buy the right stuff, you're
not lost (I have and use two SPS' here) but get a true UPS.  If you
don't get a true UPS, quit bellyaching and I'll quit preaching :-)

># Jeff Liebermann   Box 272   1540 Jackson Ave   Ben Lomond  CA  95005
># (408)336-2558 voice  (408)429-0483 digital pager      CIS:73557,2074 
># PC REPAIR & RF DESIGN  uunet!comix!jeffl  ucscc.ucsc.edu!comix!jeffl
># universe!milky_way!solar_system!earth!na!us!uunet!comix!jeffl
-- 
Bill Kennedy  usenet      {att,cs.utexas.edu,sun!daver}!ssbn!bill
              internet    bill at ssbn.WLK.COM   or attmail!ssbn!bill



More information about the Comp.unix.sysv386 mailing list