|
> my tests were done with a single real-server running flat
> out, the director
> wasn't getting much exercise at all in the VS-DR case.
> I'd have to test it with the director at high load to
> simulate Jerry's situation. I don't have enough hardware to
> do that.
>
> Joe
<snip>
http://www.linuxvirtualserver.org/Joseph.Mack/performance/single_realserver_
performance.html
Result: There is no change in throughput when only one node has a TP kernel
(TP, non-TP), (non-TP, TP) . It's not till TP is put on both kernels (TP,
TP), that throughput is affected and then only at high network load.
Presumably there is some interaction between the two TP kernels at high
network load.
</snip>
you might want to update your HOWTO performance page to indicate a little
doubt due to hardware not being able to test the throughput higher?
|
