Renderfarm.fi Work Unit policy will be tightened

Some of the late work units have been extremely expensive to compute, so much so that some of our users have been unable to finish WU's in time, leaving them frustrated. I totally relate to people's posts about the subject (see JoneK's post here) as my own mere two year old machine has been crunching some of these sessions almost continuously. Reflecting a bit, I think that until we have the technology to ..

1. Use multiple cores on the same WU (currently 1 WU / core)
2. Save render state in Blender (continue render from where the client left off)

.. we will need to aim to keep the per WU rendering time down to a maximum of 3-4 hours per core. What's a core in this case? I think a safe benchmark would be an Intel Core 2 Duo, as that particular processor has been now on the market for 2,5 years. What do you guys think?

These issues would not be such a big problem if we had the capability to effectively do preprocessing and postprocessing on the server and attach them to the WU, in which case the frames could be split into multiple part much more simply. Unfortunately this is not the case yet.

We're already working on these issues (infact Olivier's preparing a post about the new java library that we're taking into use soon) and as we move forward we'll leave these technological barriers in the past. For now however, we'll crunch the ongoing sessions to the end (so as not to make even more people frustrated) and go back to a tighter policy. Sorry for the inconvenience this has been causing you guys!