We have exactly the same problem with writing frames to server..
We have a readyNas and our workstations are Windows 7 64-bit machines.
The frame renders pretty fast but the writing the frame can take very long.
I haven´t had the time to investigate more we are in middle of production.
Please let me know if you or anybody else have any good solution.
Cheers
Daniel
Nice Ninja
-----Original Message-----
From: Dorian Fevrier
Sent: Friday, November 25, 2011 6:46 PM
To: [email protected] ; David elhijo LEROUX ; Bruno Mahe
Subject: Re: [Nuke-users] Change the write size buffer to decrease network
IO?
Hi Holger and thank for your feedback.
ITs told me the TCP ACK delay was dynamic and are not forced...
I'm sorry to not be able to tell you more... :(
Anyway, if you do some test, don't hesitate to tell me if you find
something interesting. :)
Regards,
Dorian
On 11/25/2011 05:30 PM, Holger Hummel|Celluloid VFX wrote:
Hi Dorian,
we actually observe this behaviour here, too - sometimes at least. i could
not find any conditions yet to reproduce it exactly.
our workstations are on windows (xp64 and 7), the file server is running
debian with samba which we use to connect the shares.
do you know of any similar settings like the TCP ACK delay you're
mentioning that have an impact on nuke's performance when writing files?
i'd like to check those to see if tweaking those makes any sense.
cheers,
Holger
Peter Pearson wrote:
On 23/11/11 17:30, Dorian Fevrier wrote:
Hi!
OS are Linux Opensuse 11.1 and we use NFS export.
Thanks in advance,
It depends on all sorts of things - how many NFS nodes you have, whether
NFS is configured to use UDP or TCP (and if it's the latter, the TCP ACK
delay), how busy the NFS network is, the format of the images you're
rendering out, if it's EXRs then the number of channels you're writing
out and the compression type you're using, etc, etc.
Nuke uses standard file writing POSIX APIs to write to files, and while
there will be a slight overhead in sending data over the network in
chunks (due to the TCP ACK delay) as opposed to the whole file at once
like the cp command does) it shouldn't be anywhere near what you're
seeing. It sounds very much like a network configuration issue your end,
either to do with some TCP ACK delay or fsync/flush configuration on the
filesystem that the NFS drives are using.
Peter
_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
