Re: OT: Shutdown vs Sleep

[Danny Sauer <sauer@cloudmaster.com>]

I've got some anecdotal evidence for you... :)

The workstation on my desk (which had a clean SuSE installed a few 
months ago):
dsauer@newwww:~ > uptime
   4:34pm  up 92 days,  7:31,  4 users,  load average: 1.17, 1.07, 1.06
A semi-random MacOS X box (we had a power failure 2 weeks ago and a 
concurrent building-wide UPS failure - which didn't affect the stuff on 
the sysadmin's desk):
bash-2.05a$ uptime
  4:38PM  up 13 days,  5:29, 2 users, load averages: 1.10, 1.28, 1.26

All are running dnetc clients (like everything at this office), BTW, 
which explains the 1.x load...

It's not documented outside of this post, but I've been senior sysadmin 
at this company for about 4 years now.  Within a few months of my 
starting here, I standardized the practice of leaving machines turned on 
24x7.  Out of the 60 or so machines I support here and the 15 or so 
off-site machines that stay on 24x7, I've had 4 hardware failures.  3 
were IBM 75GXP drives (which were recalled because they regularly die), 
and one was a Toshiba Laptop.  At my previous place of employment - a 
community college - my wing (about 250 nodes) subscribed to the 24x7 
machine mentality.  A couple of the other wings subscribed to the "turn 
it off at night" school of thought.  The whole school pulled hardware 
from the same supplier - Gateway.  Hardware failures were *less* 
frequent in my wing than in the power-cycling wings over the 3 years I 
spent there, and it appeared to me that we had fewer software problems. 
  The software problems could have been attributed to my mad setup 
skillz, though the hardware was identical.

I can make a graph and print it on some letterhead if you need an 
"official" source stating that 24x7 operation is good for machines.  I'm 
convinced, though.  Running at 100% load 24x7 hasn't hurt any of these 
machines, and several of them are reaching end-of-life.

--Danny

Benjamin Story wrote:
> Ok, I know this topic has been hashed and rehashed, but after a few
> googles I haven't found any firm proof one way or the other.  We'd
> like to have everyone leave their machines on at night for maintenance
> reasons, but the VP's would like firm proof (ie newsgroup flame wars
> don't count) that this won't harm the machines.  Has anyone seen any
> firm studies or research on this?
> 
> Ben Story

-
To unsubscribe, send email to majordomo@luci.org with
"unsubscribe luci-discuss" in the body.