Re: Tuning LVS

To: Andreas Jung <andreas@xxxxxxxx>
Subject: Re: Tuning LVS
Cc: <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
From: Julian Anastasov <ja@xxxxxx>
Date: Tue, 20 Nov 2001 20:00:32 +0200 (EET)

On Tue, 20 Nov 2001, Andreas Jung wrote:

> hehe. do newer versions have a better characteristic when performing
> with small packets ?

        You have to check it. You can also check without using LVS
directly with the real server(s). How many are the real serevrs? I
don't know whether these tests depend on something like TCP_NODELAY,
etc... You can try to provide some data, with gnuplot for example.
See, how Joe creates nice pictures. OTOH, I'm not sure 2.2.18 is a
good choice to test TCP. Please, make your tests with 1.0.8 + 2.2.20.
You can even try NAT with fastcsum-1.0.8-2.2.19-1.diff from

        Try to systematize you results (if that is your goal).
As for testing the throughput you can try "testlvs" from the
above URL.

> Andreas


Julian Anastasov <ja@xxxxxx>

<Prev in Thread] Current Thread [Next in Thread>