some ruminations on saving money, power, etc.

[dsr-mzpnVDyJpH4k7aNtvndDlA at public.gmane.org (Dan Ritter)]

On Wed, Sep 29, 2010 at 11:14:10PM -0400, Tom Metro wrote:
> Dan Ritter wrote:
> > - my desktop, which never sleeps
> ...
> > The...machines that don't sleep, aren't sleeping because I can't get
> > them to sleep.
> 
> Figuring out why the desktop doesn't sleep sounds like it would provide
> the best payoff for the least cost outlay. Time cost is another matter...

Agreed.

> > The server can't and shouldn't sleep...
> > Myth box don't sleep because they need to be awake. 
> 
> Maybe you can spin down the disks? (Probably only of marginal benefit.
> 10 or 15 watts down to 2 or 3 watts?) Switch to "green" (low RPM) drives
> on your next capacity upgrade?

Both plausible. I'd have to check the disk spin-up time. Some of
the Myth drives are WD "green" disks.

> How about enabling frequency scaling on the CPU? Or if your CPU doesn't
> support it, maybe it's possible to upgrade to a newer CPU, while keeping
> the remainder of the hardware. That way you leverage the most power
> savings for the least investment.

Turns out that frequency scaling doesn't actually save power on
older Athlon chips (i.e. my main server). It's in place on all
the other machines.

> Have you considered consolidating your main and MythTV servers? (I hear
> some people have even had success running MythTV in a VM.)

The idea fills me with dread. The main machine is extremely stable;
that's it's whole purpose in life: to be available. It runs mail, a web
server, and SSH. The Myth box is not particularly stable.

> > There is a timed wakeup function on some motherboards;
> 
> WOL packets sent from another box (like a router; see other message) is
> probably more manageable. But I can't see taking the effort to manually
> synchronize wakeup packets with the MythTV scheduler. To do this right
> you'd need to build some software that queries the scheduler for
> upcoming recording times, looking for gaps large enough to be worth the
> effort of shutting down, and then scheduling a wakeup task on another
> box. It wouldn't be that hard to do, and I bet someone has already built it.

Except that it tries to use the timed wakeup function, yes, it's
been built. I'm not sure I trust the machine to go on and off so
often, though, and the startup time is significant to anyone who
wants to watch tv at an unexpected time.

These are all good ideas in general, and will probably help
other folks in different situations.

-dsr-


-- 
http://tao.merseine.nu/~dsr/eula.html is hereby incorporated by reference.
You can't defend freedom by getting rid of it.