recommend a fast server?

[Gary Mc Ginty]

We are a small arch viz company and have a network of 10 PCs all accessing the same hard drive thru the network to render a job in after effects.

On a job we just completed this set us let us down. We want to configure a new system on a budget.

We want fast hard drives, fast connectability to transfer files. Any advice would be appreciated..

[SandmanNinja]

* A good, FAST, smart switch - data throughput is A1 priority

* Good, clean Cat-6 data cables - they are certified for 1000 megabits per second

* any reasonable network OS - I run my network on NT4, but I was certified on NT4

[visual3d]

i guess all you need is a raid setup...get raptor harddisk from western digital. a minimum of 2 disk is fine....if you have the money...get 4 disk..

get 2 of those terabyte harddisk as backp but not link to the raid.

[luckytohaveher]

O.K. so there is a couple of concepts people usually miss in renderfarm configuration:

 

1. Overall rendering throughput is ecosystem based. It is not all about CPU cycles, although that is critical. Before you start, go to system monitor and look at CPU load, disk load, network traffic, and memory consumed. Two issues, look to current constraints and look 18 months out or longer. You don't want to upgrade too often.

 

2. The network is important, especially the server to router connection. Get a server network card and then bridge the connection between two ports. Bridge the connect from the computers to the routers at two ports each. You want the overall throughput to be VERY fast. If two nodes call a project from the server simultaneously, you want them to both be fed at maximum speed. Be sure the network topology is star based from the server.

 

3. ABSOLUTELY, POSITIVELY, GET SUPERCACHE for the SERVER!!!! Why? Each time you load a job, frame, file, complete a render, etc. the files are called from the server to the INDIVIDUAL computer rendering. If that is disk based and shared across multiple slaves, it can kill the performance. Make sure the server has massive ram and the SUPERCACHE is configured to support the maximum size job, if possible. If you can afford it, get at least 128MB of Supercache performance and space on each of the rendering nodes. This will make the slave nodes optimized also.

 

I am sure others will have input on the CPU model, ram speed, video card, etc. Just be sure you look at the actual load, actual issues, and actual demands. After all, it is an ecosystem -- regardless of what the Intel/AMD marketing machine would want you to believe.