Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 9 Sep 2004 00:24:47 +0400
From:      Gleb Smirnoff <glebius@freebsd.org>
To:        Julian Elischer <julian@elischer.org>
Cc:        net@freebsd.org
Subject:   Re: [TEST/REVIEW] Netflow implementation
Message-ID:  <20040908202447.GA5179@cell.sick.ru>
In-Reply-To: <413F4BBE.1020304@elischer.org>
References:  <20040905121111.GA78276@cell.sick.ru> <20040908103529.V97761@murphy.imp.ch> <20040908085607.GG597@cell.sick.ru> <413F4BBE.1020304@elischer.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Sep 08, 2004 at 11:13:18AM -0700, Julian Elischer wrote:
J> >This is working solution, but not correct. :)
J> >To catch both directions you should feed ng_netflow with incoming traffic
J> >from all interfaces.
J> >
J> 
J> using 'tee' means you are duplicating all packets.
J> shouldn't you do collection "inline? or does this NEED to have copies of 
J> the packets?

This is in my TODO and TOSEE list. I'm not yet sure that this would be better.
There are some advantages in current state: packets are processed with no delay,
and a copy is queued for netflow processing. In case of multiple interfaces
attached to netflow node we can serve them simultaneosly, without waiting for
lock on single netflow node.

A good solution would be to send only IP and TCP header towards netflow node.
Is there a standard way to do this?


-- 
Totus tuus, Glebius.
GLEBIUS-RIPN GLEB-RIPE



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040908202447.GA5179>