[mythtv-users] A clustered PVR just rocks...

lemongecko at gmail.com lemongecko at gmail.com
Fri Jan 18 15:14:33 UTC 2008

On 1/18/08, Matt Emmott <memmott at gmail.com> wrote:
> On Jan 18, 2008 8:44 AM, Steve Smith <st3v3.sm1th at gmail.com> wrote:
> My home network right now is 1 myth FE/BE, one FC6 fileserver, One Windows
> domain controller / fileserver, my Core2Duo Vista workstation and a couple
> other Windows PCs. It would be awesome if I could distribute the CPU load
> across some of these machines.

You could always look at a system like Condor (
http://www.cs.wisc.edu/condor ). It's a batch queueing system that was
created at the University of Wisconsin for use originally for helping
out their computer aided engineering. At work, we use it as an
impromptu compute grid to allow our researchers to catch up on
smaller, non-parallel modelling tasks if the high-performance cluster
is booked up.

The nice thing about it is that, as its name kind of implies, it's a
scavenger. The Condor daemons run in the background, waiting until the
machine is sensed to be idle. It then announces its availability for
running jobs to the central condor master. With a little shell
scripting, one could easily run the Condor submit as a user job. At
that point, any jobs waiting in the condor queue can be sent to that
free machine. If you have binaries for both Windows and Linux
(mencoder comes to mind), you can use both types of machines. It's
smart enough to know if you have a multi-cpu / multi-core run host,
and the run host will announce itself as  separate virtual machines,

At night, one wing of our building turns into a 251 TFLOP workhorse.

It can be a bit of a pain to set up and you have to be careful how a
machine defines itself as "busy" (because it'll evict a running job if
a machine gets taken back over by the user), but I can offer some
fairly generic configs for Linux and Windows hosts if someone's truely

More information about the mythtv-users mailing list