-
Cluster rendering: what’s wrong?
[First of all, sorry for the 2 posts in a row. Is there any netiquette rule in this board about not posting too many separate messages, even if they are about completely unrelated subjects?]
We did the other day a cluster rendering test. We have several dual G5s with 1.5 Gb. of RAM, linked together with 100Mbp Ethernet, and we configured 3 of them in an unmanaged QuickCluster (without touching anything, just leaving the default options). We sent a job to the cluster (a conversion from a FCP sequence to MPEG2 using Compressor) that takes usually 40 min. for one computer to render… and the 3 computers together took 51 minutes. Is this normal? Could the bottleneck be in the 100 Mbps. Ethernet?
(Some additional notes: we’re using FCP 5.0.4 and Qmaster 2. The 3 computers weren’t running anything else that was CPU-intensive; in fact, according to the Activity Monitor, none of the 2 other computers (the ones that weren’t the master) were having more than 10% load).