Sharing repository on remote farm

Deadline
1

Hi !

We are about to do a setup and I just want to double check with anyone here if this is a good idea:

We want to VPN-share our deadline repository on a remote network, so 3 remote slaves can connect to it and start render the job. So few things:

  • We dont want the remote slaves to save the frames through VPN directly here, we are rebuilding the project architecture locally on the remote machines and use a sync system to get the files. Because we do that, we will have to ‘mount’ the deadline repository on another drive than the production drive. So the repository Path will not be the same on the remote machines than on our local machines. I guess this is not a problem ?

  • Will we have a sync issue ? Remote machines are in Poland and we are in Canada so we have to expect a certain lag in the communication. Is this a problem ?

  • Any other advice or flags it might raise when you read this ?

Thank you very much !
Gab

2

Hi Gab,

I suggest using the Path Mapping feature for your remote slaves so that they can can read inputs and write outputs from the remote storage.

With regard to syncing, you could populate the Assets in the Dependencies section of each Job. This will prevent the remote Slaves from rendering until the assets are in the remote storage (since the path mapping will apply to assets as well).

That should get you started. If you have other questions or issues as you go, just let us know.

3

Hi ! Thanks for answering

I dont want the remote application to source and write remotely. The reason is that i have about 25000 frames to render, and they take 3 minutes each. so read/write time is an issue

I prefer having all the source files local and render local, and then get the rendered frames here via another process, while the app is rendering the next one !

4

What I meant by “remote storage” is storage (NAS/SAN) that is local to the remote Slaves. Path Mapping + Required Assets will mean that the remote Slaves look for the assets in the storage that is local to them but not start Jobs until the assets exist there. Likewise the render output would go to the remote NAS/SAN and you can then pull down the results as needed. However, if you have geographically-distributed Slaves working on the same Job, it means that the output for the Job will be geographically scattered, so that’s another consideration for the design of the overall pipeline.