We had some changes to a piece I compressed last Friday and had to recompress the updated version in a hurry today, so I set up a quick cluster with a dual 2.6GHz G5 controller node and 2 QuadCore G5s. I followed the manual and it was set up in about 5 minutes.
I compressed a 4 minute DVCPRO HD FCP clip to a 1920×1080 MPEG-2 multiplex’ed file using compressor and it took 20 minutes. When I compressed the same file on Friday from the dual 2.6 G5 it took 45 minutes. I don’t know what I did right or differently, but it worked as advertised. I set the networked computers as ‘service only’. We may be gigabit ethernet, but maybe not, since we’re tied into the corporate LAN as well. I mounted a common shared drive drive with the original media on all three systems – but I don’t think it will work anyway unless you do this.
I initially thought it was taking longer than the original local compression because the status read 3% complete with nearly 2 hours remaining – until I switched the batch monitor window to show the cluster I had set up and realized it had burned through 90% of the compression in 15 minutes (there’s a drop-down window on the top right of the batch monitor window in compressor where you can check the status of each computer in the cluster including the local controller). Thinking back, I think I paused the compression on the controller machine and double-checked the service machines and then came back to the controller and started it again. Maybe this kicked the other processors up a notch or something, but I really don’t know. It’s discouraging to hear about the mixed results you’re all having, though. I thought I was really on to something.
I tested the file and it looks great – and done in almost a third of the time. I was pretty impressed. I’m now looking for a way to harness Qmaster for AE renders. Any pointers? I don’t want to write custom command lines. I just want to send the render to the Quick Cluster I’ve set up. Probably wishful thinking, though.