> However, at -very- high packet loads it is far less CPU-efficient than
> getting the ARP settings correctly working. The REDIRECT
> method was bogging
> down our LVS boxes during peak traffic, something which does
> not happen with
> doing it the 'right way' with LVS-DR and silent arp-less
> interfaces on the real
> servers.
I disagree. If you look at Joseph Mack's throughput analyasis of the
different methods, you'll see that efficiency is effected only when you use
two redirects. I suppose if you had TONS of small packets (abnormal
amounts) then perhaps there might be an effect.
I would be interested in seeing some kind of actual performance indication
if you have proof otherwise, since I am running the horms method here.
gracias & good weekend to you all.
Peter
|