Hi LQ, this is my first post, and I'm not totally sure that this is the right place to write.

I would like to build a linux cluster made of 4 computers, that will see all the resources as a single machine with many cores and a lot of ram. I would like to run KVM with several virtual machines and I would like to gave to some of them more CPU than a single computer has.

Do you think is it possible? Are there any preferred configuration to do this? Are there any specific software to get this result?

Thank you in advance and sorry for my english, I'm italian.

Luca

There are ways to do HA (High Availability) with KVM, and there are tutorials. But, such things are NOT using multiple computers as a 'single entity'. They are more for high availability and load distribution. Hadoop or Gearman is probably the best way to go to get started:
http://hadoop.apache.org/
http://gearman.org/

Be aware, though, that while this is a cool CONCEPT, and it does work...not everything is going to work as-is. The code you submit to such clusters has to be WRITTEN/modified, to employ those resources. Which is sometimes harder than you may think.
Your english is pretty darned good. While I'd be happy to converse in Italian, this is an English-only forum.

Check out Open Grid scheduler. It is the open source version of Sun's N1 Grid Engine

http://gridscheduler.sourceforge.net/

I am currently running it in a cluster of 8 nodes and it works pretty good.

But as TBOne said all of the software has to be written/modified to use the OpenMPI API.

Not to difficult and there are lots of examples out on the net.