Baslinux mailing list

[qwms-avib AT dea.spamcon.org]

James Miller wrote:
>
> the computing power of all machines hooked to a given 
> network such that this power can be used from any of 
> the "nodes" (correct terminology?) on the network: is 
> that approximately correct?  

Yes, with some limitations.  Many processes won't migrate
to other nodes, because they are too short (and therefore
not worth the time it takes to migrate) or because they
are written in a way that ties them to the one system.

> If so, then BL2 + Openmosix = a way of utilizing older 
> (maybe throwaway) systems to get more up-to-date computing 
> power, right?  

Yes, the effect would probably be most noticeable on a network
of old PCs (because the slower a system is, the more worthwhile
it becomes to spend the time migrating the process).  Of course,
a slow network (10 mbps) increases the migration time and makes 
migration less likely.  Fortunately openMosix does all these 
calculations automatically and quickly decides whether migration 
is worthwhile.

> A person could salvage say, 3 or 4 older machines and, with a 
> bit of networking setup, have the equivalent of a much better 
> machine than any of the individual machines on the network.  

Yes, this would be most noticeable when doing lengthy processes,
like a kernel compile or SETI@home.  However, with two or more old 
PCs, it would still be worthwhile to use the fastest one to run the 
programs and a different one to run the Xserver.  Combined with a 
well-tuned openMosix, that would give you the best performance.

> Strikes me as a great way of keeping obsolete equipment out 
> of the landfill, 

It works even better if you take one up-to-date computer and
network it to a dozen old PCs (which are run as Xterminals).
I see that the LTSP is now including openMosix as an option.

Cheers,
Steven