Hi all again :))

> I will take that into consideration. I will have to think about how web
> serving and proxying are alike and different.

Actually they are not very similar, but still LVS is better. I try to
explain myself:

* Open Mosix works well when it doesnd need a local, unique storage. In that
case, forking the process to a different computer is bad as you take the
process away from the data. In case of proxying this is NOT a problem as all
cache is valid, no matter where it is (be it the original or the forked CPU).
So it might seem OM is OK for this.

* BUT, in OM you imposse a load managing the processes. The processes move
around based on which CPU has some spare power. This adds some latency
(process jumping, management, algorithms) and consumes CPU by itself.
Instead, with LVS you have a defined number of processes in a cluster of
CPUs, and the LVS daemon just keeps sending data to each one of them, it
doesnt FORK a process (that by itself consumes CPU) but sends data to a
different daemon. If you consider the same number of DG processes in all CPUs
in both cases, LVS is much lighter to the system as it doesnt need to fork,
nor control where a process is, or if it needs to come back or jump again.

* Actually I think DG by itself is a good analogy. As you probably know, DG
usually worked creating a set of processes and as was needed, forking new
ones. This was very heavy on CPU with high loads. In recent releases you have
the choice of preforking a big number of processes and just send them data.
This might consume more CPU in low stress situations, but sure as heck is
better when things starts to climb. Some people report more than a 50%
reduction in CPU usage !!! With OM and LVS you have more or less the same
concept. Wuawww, that was a good analogy !! I clap myself :)))

So in essence, the apache analogy was not a good one, but still, IMHO, LVS
is less stressfull for a cluster. But heck, if you want to try, I can even
give you David's email and you can comment the results to himself :)

Hope this one definetly helps :)

--
Jaime Nebrera - jnebrera@...