On Tue, Sep 19, 2000 at 10:42:48AM -0400, Joseph Mack wrote:
> A while ago someone posted a link to some mrtg/iptraff type gifs showing
> the cpu/disk/IO for 4 identical real-servers for several hours. The graphs
> for the real-servers were indistinguishable.
>
> What are you going to do if you find that (say) disk activity is too high
> on one real-server? All other real-servers are going to be having equally
> busy disks.
Firstly, no idea. Well...
Secondly, that's good news though... because it means that LVS is doing its job
well.
Thirdly, I think... if this is the case (everything being equal), then
something I thought about a while ago might be relevant.
And that is to automagically (again a simulation is good here) wake up
additional real servers to help with the load -- ie. add them to the farm
dynamically. This would be cool.
Of course then one has to wonder why are they in reserve and not all running? I
don't know. Hmm...
--
B. http://b.makelinux.org/ "Of course there is a reason why!"
__________________________________________________________________________
If you live long enough, you'll see that every victory turns into a defeat.
-- Simone de Beauvoir
|