[mythtv-users] NFS and remote backend

[Patrick Ouellette]

On Mon, Jan 14, 2008 at 05:21:24PM -0500, Ryan Steffes wrote:
> 
> > The other programs using the network may be more forgiving of timing
> > issues.  The display of video without stuttering requires the information
> > for the video frame to get to the frontend with enough time to be decoded
> > and rendered. If the frontend is waiting on data from the network it
> > will stutter.
> >
> 
> Well that's what I mean.  I can bog down the network writing any
> number of files and reading files with no stuttering, as long as it
> isn't the file that's being recorded right then.
> 
> In other worse:
> 
> PVR-150 can record show A while I watch show B and it will be fine.
> PVR-150 can record show A while I watch show A from NFS and it will stutter.
> PVR-150 can record show A while I watch show A streamed from back end
> and it will be fine.
> 
> Ditto for other tuners.  I can record several shows from both PVR-150
> and my A180 and watch something and it'll be fine, unless the
> something is being recorded right at that moment.  That makes me think
> it's not the network specifically, but something about the network set
> up, like a refresh issue.
> 

OK, guess I was being a bit dense.  If you ONLY have problems watching
the show that is currently being recorded *when* it is being recorded,
and after the recording finishes you can watch the show without
problems, then I would say your network is not the issue (and NFS is not
the issue).  I'd look to MySql tuning or a disk cache related issue.

> > You can try the following to see if anything stands out as odd:
> >
> > Run some NFS tests with large file reads and writes on the NFS share
> > from the 100 mega-bit frontend.
> >
> > Change your NFS block size on the client & server (restart NFS) and run
> > the tests again.  (large file means many times greater than the 32K
> > block size)
> >
> > Repeat the above tests with a crossover ether net cable connecting your
> > server and client (at 100 mega-bit). This eliminates the D-Link
> > router/switch.
> >
> > Install SAMBA, and run the tests again using the CIFS share from SAMBA.
> >
> 
> Can give that a shot, I don't know of a particular reason to not
> stream from the backend, since it seems to work better.  I just kinda
> wanted to know /why/ it worked better.
> 

I've always been one to go with what works for me.  If streaming from
the backend works, that's great. 

Be interesting to know what the difference is.  Guess a trip into the 
SOURCE is required (unless a dev wants to chime in).  I bet it has to do
with when the backend is streaming the data it doesn't have to wait for
a disk write, but when the data is not streamed but shared the client
has to wait for the backend to complete the disk write. (Pure conjecture
on my part.)

Please let me know how it turns out if you continue to investigate this
issue.

Pat
-- 

Patrick Ouellette                 pat at flying-gecko.net
kb8pym at arrl.net                   Amateur Radio: KB8PYM 
Living life to a Jimmy Buffett soundtrack
"Crank the amp to 11, this needs more cowbell - and a llama wouldn't hurt either"