Compressor 4.1.2 Clusters & Distributed Processing Problems
So I must be doing something wrong in setting up several computers as a rendering cluster for Compressor 4.1.2. Here is my setup. I am trying to include every single bit of what may be important information:
Gigabit Hub: Trendnet TG-S80G Giagbit switch. All Cat6 Cable running to all computers, WiFi turned off.
Main Computer: iMac 12,2 (27 inch, 4 core 3.4 ghz i7), 20GB Ram (have two thunderbolt RAID arrays, a Pegasus 12TB as the main capture / storage array and Lacie 10TB as the backup)
Computers hooked to hub to be used as rendering clusters:
Quantity 2 11,2 iMacs (21.5 inch, 2 core i3 3.4 ghz 4GB Ram)
Quanitity 3 12,2 iMacs (21.5 inch, 4 core i5 2.7ghz, 4GB Ram)
All have mechanical hard drives. All computers are running 10.9.4 and Compressor 4.1.2 All have Compressor open and running, set to share, none of the 21 inch iMacs will allow “Enable additional instances of Compressor”, all have sharing on, have "ethernet" as the choice under how to share resources, and on main computer in Compressor two new clusters were created: one that included all computers on the hub including the source computer, and one that did not. Cannot adjust any of the iMacs to do jumbo packets (apparently a limitation of the chipset). File sharing, screen sharing, and remote login are all enabled on all computers (since I have students working on them during the year I like to be able to remotely look at their screen and see what they're doing from my rMBP)
A 50GB 1080/24p ProRes file (about an hour, entire act of a play our school performed) compressed into SD for Apple TV Format Compressor present. Compressor open and running on all computers. ProRes file is taken from the Pegasus RAID array and written onto mechanical internal storage on the 27” iMac.
Using “This Computer” and the entire cluster: 21 minutes 26 seconds:
Using “This Computer”: 23 minutes 42 seconds:
Using cluster without “This Computer”: 59 minutes 17 seconds.
On switch all lights are flashing indicating network activity (and nothing else was being done on them, no downloading or anything), so it must be Compressor traffic. The difference in crunch time with or without the cluster is obviously negligible.
So what the heck am I doing wrong?