Beowulf Cluster for DC?

[superkdogg]

I'm thinking of trying something new and combining a few of my old rigs into a 3-4 machine Beowulf-style cluster. Being also a Linux n00b, this will be quite a challenge for me, but nothing ventured=nothing gained, right?

Anyhow, my question is: Does anybody know if the overhead of splitting the workload would take away from the overall productivity of the machines? Is it possible that each machine's individual overhead is greater than the total amount in a Beowulf and that the production could increase?

LMK what you think.

Has anybody done this before?

 

[bluestrobe]

I asked about this about a month ago and got a few similar responses. Unless the program being ran is setup and coded for dual processors, the cluster node would be a waste. Looking at some websites about it showed the same thing too.

 

[superkdogg]

Originally posted by: bluestrobe
I asked about this about a month ago and got a few similar responses. Unless the program being ran is setup and coded for dual processors, the cluster node would be a waste. Looking at some websites about it showed the same thing too.

What about with load balancing software[MOSIX]? What if it iscoded for multithreading?

 

[bluestrobe]

On the linux side, I have no answers.

 

[InlineFive]

I would run each machine individually. Connections at a high enough speed between those computers to act as a cluseter would make it expensive and impractical.