Jacob Joseph wrote:
> Hi. While it isn't possible to distribute chunks of NTSC media, is
> there any easy way to use cluster mode to distribute entire jobs as a
> single chunk? This would greatly simplify utilization of multiple
> machines from a single queue. Sure, it's possible to use the resulting
> transcode commands manually on the target machines, as referenced in
> http://www.exit1.org/archive/dvdrip-users/2002-09/msg00139.html, but it
> seems that such functionality could easily be provided. As that thread
> is quite old, has this already been done? Can anyone point me to some
> manner of workaround?
Try setting the chunk size greater than the movie's number of frames,
but I'm not sure if cluster scheduler applies a minimal chunking
nevertheless, because it tries to parallelize transcoding of single
projects which usually needs at least two chunks. But you could give it
try, if you didn't already ;)
Another solution is the brand-new (since 0.97.3) dvdrip-tet program,
which is a command-line dvd::rip transcoding frontend. If you take care
that all your network nodes have the same file system layout of the
dvd::rip project directories (resp. the absolute paths must be
identical; use according NFS mount point symlinks to accomplish that)
you can start one dvdrip-tet program for each project on its own node.
Apart from that I'll verify the chunk-size-increase-hint and will make
it possible, if it doesn't work yet. So in future you should be able to
do that even with the "standard" cluster mode mechanism.
Regards,
Joern
--
LINUX - Linux Is Not gnU linuX
pgp00023.pgp
Description: PGP signature
|
DOCUMENTATION
TABLE OF CONTENTS