Hey all, I’ve been working on a simple but endlessly versatile render farm manager.
The goal of SARFIS Pro is to let users set up a render farm of their own over LAN and or WAN with minimal effort.
Easily handle multiple 1000s of machines.
Very low CPU / RAM usage.
Sub millisecond request times (LAN)
The ability to run tasks with any software not just Blender.
Node based task designer.
We have been using older versions of SARFIS in production for years with over 2.5k machines online at times and with single jobs over 100k tasks/frames so we decided to polish it up and release it to the public.
At the moment we have a closed beta with future plans to allow people to buy SARFIS Pro for very very cheap. Just a few dollars for people who just want to use it on their home machines. Of course we will have larger plans but my goal is to help every day blender users not just studios / companies
Here are a few pictures of our work in progress GUI, no nodes yet but they are coming very soon!
I missed this! It looks like a really interesting project, last year I was in dire need of something like this. Now I am not messing with render farms so much, but I’ll keep an eye on this none the less. Unfortunately I don’t have much hardware at my disposal, so I am not sure how much testing I could do
@Bernardo if ya ever need to do more rendering just let us know By day I optimize cycles scenes beyond belief and by night I build render farms and render farm software.
We have been making plenty of progress on the GUI. We built a new file server and optimized the queue manager a bit more, we believe we could easily get 20,000 machines running with the pre existing queue manager. I have tested the solution with ~4000 machines and it handles it like a dream. Next I will test on 8000 machines!
We have been trying to figure out the best way to manage logs. Blender spits out 20+ MB logs when you have small tiles and large frames. We are considering compression or building a system that doesn’t repeat similar lines of logs only the changed values. To ease the pain this very moment you need to enable logging on the machine explicitly (its just a check box) this way when we run 4000 machines we don’t have 80GB of logs
We refactored the file server it is now follows standard http url paths. You could technically use the file server in your own projects because it has a basic API
We added the node system, it allows you to run any EXE or Python script on the remote machine. We have a few other nodes to download / upload files easily from the Sarfis File Server and some for Google Cloud Storage.
We added inline expressions to the nodes so you can do math, regex or use variables (as seen on the left)
wow this looks quite awesome! At the moment my work doesn’t require anymore the use of a render farm, but I’ll definitely keep this under my radar! Keep up the great work!
By day I optimize cycles scenes beyond belief and by night I build render farms and render farm software.
