Author Topic: Sharing the resources of the render farm  (Read 4908 times)


  • Full Member
  • ***
  • Posts: 13
Sharing the resources of the render farm
« on: November 12, 2007, 08:50:04 AM »

We are a school who don't have developers and we are planning to use Qube to manage our render farm which we use for 3dsmax and Maya. We are trying to share the resources of the render farm between different student's projects.

For example, let's say that our render farm has 24 worker hosts. We would like that, if two students (which we can either authenticated by IP address or Windows login) submit jobs, the job of student A will get half of the workers (12 workers) and the job of student B will get the other 12 workers. We would also like that if a third student C submits a job, then his job gets a third of the render farm (8 workers) and the jobs of students A et B are automatically reduced to 8 workers each.

Is there a simple way to achieve this automatically? (i.e. without requiring the students to type a priority or a cluster name when submitting the job, because they might cheat on that)

We also have another situation we would like to handle, it's when a student is late for his project, we would like to be able to make Qube always treats his jobs as top priority, regardless of already existing jobs. Is that possible?

Thank you


  • Hero Member
  • *****
  • Posts: 229
Re: Sharing the resources of the render farm
« Reply #1 on: December 12, 2007, 08:19:00 PM »
On the first point, you are looking for a autobalancing queuing algorithm sometimes referred to as "fair share." Currently, we don't offer such a queuing algorithm.

As an alternative, You could partition your farm into a number of even-sized clusters, each assigned to a student. The students would still have to put in their cluster, unless you were willing to modify the Qube GUI to submit with a cluster based upon the student ID. Modification of the Qube GUI would require comfort with Python.

In the second point, you could have the student submit with a very high numerical priority.