sdarpel
02-15-2005, 08:18 AM
Anyone and everyone, I am at my wits end. I have a major project to render for my animator, and have been working with various network rendering setups.
First, I played with Screamernet II, but was very disappointed with how inefficient it was as a manager. My nodes, anywhere between 6-12, spent most of their time waiting for work, not rendering. This was on an XP Pro box.
Next, I put SN II on a Win 2003 Server box, but the performance was not much better. Plus, setup and running on a pure SN II net was a pain. Specifically, having to go to each machine and launch the nodes was a pain.
Enter Tequila. At first, I was very pleased. The ability to remotely launch, kill, and restart nodes, along with the management functions for the scenes was very nice. I started with a small scale test, including 6-7 processors, all inside my office. It performed fairly well on a 1350 frame animation.
Things began to fall apart once I started up a full scale test. I started to render on about 20-25 machines, mostly all P4, 3.0 GHz HT, 512MB-1GB Ram. My server kept cutting out. Granted, it was an old Dell Poweredge 1300. So, I seconded one of the desktops to a Win 2003 Server machine and ran on it. Same problem. So, I figured, well, desktops aren’t meant to run as servers. Finally, I put the whole shebang on a Dell Poweredge 4400, with Dual 1.0 GHz Xeon Processors, and dual 1 gigbit NIC’s. I set it up to control 16 new (new to me, they are all Athlon 600 MHz’s) pc. It ran great from 1:30 to 6:30 pm fine. Then, BANG, same issue, and it essentially took my server down.
PLEASE, PLEASE, PLEASE, can someone tell me what to use and how to set up a RELIABLE network render farm, where I can render out at night and on weekends on the 30+ pc’s I have available???? The whole point is not to have to baby-sit this **** thing. I have a huge render to do in about two weeks that may very well determine our funding situation for the next 10 years.
Thanks,
Scott Darpel
You can reply directly if you wish to scott@darpel.com
First, I played with Screamernet II, but was very disappointed with how inefficient it was as a manager. My nodes, anywhere between 6-12, spent most of their time waiting for work, not rendering. This was on an XP Pro box.
Next, I put SN II on a Win 2003 Server box, but the performance was not much better. Plus, setup and running on a pure SN II net was a pain. Specifically, having to go to each machine and launch the nodes was a pain.
Enter Tequila. At first, I was very pleased. The ability to remotely launch, kill, and restart nodes, along with the management functions for the scenes was very nice. I started with a small scale test, including 6-7 processors, all inside my office. It performed fairly well on a 1350 frame animation.
Things began to fall apart once I started up a full scale test. I started to render on about 20-25 machines, mostly all P4, 3.0 GHz HT, 512MB-1GB Ram. My server kept cutting out. Granted, it was an old Dell Poweredge 1300. So, I seconded one of the desktops to a Win 2003 Server machine and ran on it. Same problem. So, I figured, well, desktops aren’t meant to run as servers. Finally, I put the whole shebang on a Dell Poweredge 4400, with Dual 1.0 GHz Xeon Processors, and dual 1 gigbit NIC’s. I set it up to control 16 new (new to me, they are all Athlon 600 MHz’s) pc. It ran great from 1:30 to 6:30 pm fine. Then, BANG, same issue, and it essentially took my server down.
PLEASE, PLEASE, PLEASE, can someone tell me what to use and how to set up a RELIABLE network render farm, where I can render out at night and on weekends on the 30+ pc’s I have available???? The whole point is not to have to baby-sit this **** thing. I have a huge render to do in about two weeks that may very well determine our funding situation for the next 10 years.
Thanks,
Scott Darpel
You can reply directly if you wish to scott@darpel.com