clustering difference between lvs and mosix

["Joseph R. Hoot" <joe@xxxxxxxxxxxxxxxxxx>]

I understand the differences between Beowulf and Mosix.  With Beowulf, each process of a "cluster compiled" application can be load balanced across all clustered servers.  With Mosix, each process of an application or thread will be executed on the cluster node that is least used (Has resources available).  In these types of situations, where does LVS fit in?  Does an application have to be compiled to use header files from LVS (aka PVM style)?  Is it application independant and more kernel clustering?

The best situation that I would be looking for is where I wouldn't have to recompile any applications and that it would load balance applications (single processes) across multiple nodes.

I've already decided to go with Mosix unless LVS does a similar thing, in which case I will probably try LVS instead.

THanks for any help.
Joe