[6183] in Release_7.7_team
Re: Athena Release Team 1/21/09 Notes
daemon@ATHENA.MIT.EDU (Jonathon Weiss)
Fri Jan 23 16:24:21 2009
Message-Id: <200901232123.n0NLNWpQ022574@speaker-for-the-dead.mit.edu>
From: Jonathon Weiss <jweiss@MIT.EDU>
To: Jonathan Reed <jdreed@MIT.EDU>
cc: Mitchell E Berger <mitchb@MIT.EDU>, "andrew m. boardman" <amb@MIT.EDU>,
Bill Cattey <wdc@MIT.EDU>, release-team@MIT.EDU
In-reply-to: Your message of "Thu, 22 Jan 2009 18:08:36 EST."
<D2194E7D-B1C3-401D-AC5D-87B831AEC9F6@mit.edu>
Date: Fri, 23 Jan 2009 16:23:32 -0500
X-Spam-Flag: NO
X-Spam-Score: 0.00
I did a quick walk through a couple of clusters today since I was in
them for other reasons. Yes, it is harder to spot down machines.
However, there are a couple of things that you can look for that will
catch a lot of them. 1) as Jon notes, look at the power LED on the
machine. 2) If the machine has a blank screen, look at the LED on the
monitor (if it does not indicate power save mode, that is a problem)
3) look for the LED's on the keyboard (if they are blinking that's
also a problem, and one that you wouldn't necessarily have caught by
looking for the athena idle screen).
Jonathon
> *shrug*.
>
> I made the rounds of most of the main group clusters today, and
> investigated a bunch of machines. Interestingly enough, the vast
> majority of broken machines in W20 seem to be WIN.MIT.EDU machines,
> but that's another issue.
>
> Once I began training myself to look at the power LED and not the LCD,
> I find it's not a big deal. Of the downed machines I found, some
> were stuck in BIOS errors where the screen isn't off, and others were
> just plain off. I found one machine that had screenblanked and had an
> error, but it was in console mode (caused by unplugged network cable)
> and console mode screenblanking was already enabled before this change
> (or rather, we never disabled it).
>
> The calculations were made based on the standby power consumption vs
> active power consumption, based on numbers from the LCD manufacturers
> websites. The Dell 2009WFP consumes 70W in active mode and 2W in
> standby mode. That's a huge difference. I mean, it's possible
> they're making the numbers up, but it seems unikely.
>
> Before the change, we debated whether or not we needed to notify end
> users of the change. The consensus was "Wait and see". It sounds
> like this is an argument for putting up some posters, telling users to
> press a key, check the power LED, and if all else fails, e-mail hotline.
>
> -Jon
>
> On Jan 22, 2009, at 5:34 PM, Mitchell E Berger wrote:
>
> >> I just noticed this, which I don't remember discussing but may have
> >> forgotten:
> >>
> >>> We deployed the power saving code. No complaints so far.
> >>
> >> My initial anecdotal experience is poor; on Tuesday, I sat down in
> >> front
> >> of a machine in 1-115 which turned out to need some help. I've also
> >> heard verbal complaints from people who would once have noticed and
> >> potentially fixed borked cluster machines but now can't.
> >>
> >> I'm currently unconvinced of the value of this change. Is the energy
> >> savings worth the extra time hotline is going to need to spend on
> >> maintenance? (I realize it's also a done deal and unlikely to be
> >> readdressed. But if you're soliciting complaints, well, here you
> >> go.)
> >
> > Likewise. There would have been a major difference in energy
> > consumption
> > back in the days of all CRTs, but now that we're so LCD heavy, I think
> > we've given up more than we're gaining by doing this. I certainly
> > used
> > to pass through clusters when I had a couple minutes and fix
> > machines to
> > save hotline time (and users time waiting for hotline to notice), and
> > now I can't... or at least I'm certainly not going to walk around and
> > try to wake up each machine to see if it's okay.
> >
> > Also, I realize that any savings is theoretically good, but in the
> > grand
> > scheme of things, $10k a year? We've adjusted all Athena machines
> > for what
> > would've allowed us to renew just 4 or 5 more boxes at the end of the
> > year? I think we'd have a larger number of usable machines if we
> > saved
> > the $10k by buying 4 or 5 fewer new machines a year and instead
> > fixed the
> > larger number of machines that we now don't realize are down.
> >
> > I know, I know, it's not Green.
> >
> > Mitch
> >
>