There isn't really a built-in capability for prepping a host like you've described. The Qube model does presume that you have the capacity and bandwidth to support for each Worker direct access of network storage.
You would want to look to one of two strategies for dealing with this issue:
- Running a prep job that copies the files, followed by a dependent job that does the render. The difficulty lies in making sure that all the Workers are prepped (the host_list flag will run a single job all hosts on the farm, or a specified list of hosts) so that the jobs could run correctly.
- The second approach would require modifying the backend of the Job Type (located in jobtypes/typename so that the initialization phase includes a copy of the file. Cleanup would happen in the cleanup stage at the end of the job. This wouldn't be too hard a modification as the code is well-structured.
That being said, I'm not sure that copying the files really buys you anything over direct access. If you have all the Workers copying the file, you'd probably consume the same bandwidth anyway.