AWS Thinkbox Discussion Forums

Feature requests - Frames per task per slave/C4D DR

Deprecated Deadline 6 Beta Discussion
#1

  1. Sometimes there are slower slaves mixed in with faster machines on a farm and often times you have to set the “lowest common denominator” so you’re not stuck waiting for slow machines at the very end of a render. Would be great to have some kind of “frames per task” override so you could force slower slaves to have smaller batch sizes while the faster machines have bigger ones. Right now it’s always a question of how big a batch can you do (more efficient for launch times) vs. if they will be laggers toward the end. I know you can make pools out of the faster machines but you’d still be waiting if the slower machines picked up the end frames anyway.

  2. Been using C4D’s Team Render for single frame previews and it’s been great so far. Would be good to have some sort of feature that TR or Deadline could talk to each other so you don’t overload a machine if it’s already DL rendering and a TR preview were to be submitted.

#2

  1. This isn’t something that’s currently possible with Deadline’s architecture because of the way task objects are stored. A task is treated as a single unit of work, and can’t be split up on the fly. What are your frames per task on average, as well as your average render time per frame? If frames are taking 5 minutes or more, you’re probably better off just setting Frames Per Task to 1. Yes, there is a bit of overhead between tasks, but there is also less risk because an entire chunk won’t have to be rendered again if an error occurs mid-render.

  2. It’s on the wish list. We’d like to look at doing a system similar to our VRay DBR support:
    thinkboxsoftware.com/deadline-6-vraydbr/

Cheers,
Ryan

#3

  1. It was mostly for After Effects renders when doing a long sequence. Most of the time the machines were fine at a batch of 10 or 20 but at the very end the comp had a lot of big layers/effects and the slow machines would take upwards of 20+ mins/per task. I ended up requeing the ending as a separate submission with single frames but since it was pretty consistent I wanted to know if there were a way to assign it per slave instead of per job. Like a ‘slow machines’ pool that only did single frames. Guess I’ll have to stick to the methods I used before.

  2. Awesome.

Privacy | Site terms | Cookie preferences