[47204] in North American Network Operators' Group
Re: Large ISPs doing NAT?
daemon@ATHENA.MIT.EDU (Marshall Eubanks)
Mon Apr 29 11:54:25 2002
From: "Marshall Eubanks" <tme@multicasttech.com>
To: Beckmeyer <beck@pacbell.net>, nanog@merit.edu
Date: Mon, 29 Apr 2002 11:53:52 -0400
Message-ID: <web-1347468@multicasttech.com>
In-Reply-To: <3CCD6A0F.3050406@pacbell.net>
MIME-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 8bit
Errors-To: owner-nanog-outgoing@merit.edu
On Mon, 29 Apr 2002 08:43:11 -0700
Beckmeyer <beck@pacbell.net> wrote:
>
> Is anybody here doing NAT for their customers?
>
> I'm looking at a situation where I may have to provide
> NAPT for tens of
> thousands of users and am curious as to what hardware is
> being used, how
> well it scales, what kind of loads it takes such as:
>
> throughput,
> max simultaneous sessions experienced,
> session establishment rates,
> avg # of sessions per user,
> ALGs you've found necessary,
> number of sessions supported per public realm IP in
> reality.
>
> I've done a survey of firewall, switch, and router
> companies so I have
> their reported numbers and I've done a bit of testing in
> my lab and have
> found that reported numbers do not necessarily translate
> into what the
> box will experience in something resembling a production
> network. This
> is why I'm asking this group - reality can bite!
>
> A second area of concern I have is how to enforce AUPs
> when your users
> "appearance" can be *very* transitive making tracking
> back the offender
> nearly impossible.
>
> Any small piece of help, advice, or pointer would be most
> appreciated.
>
> Thanks most much.
>
Is the whole problem just a lack of address space, or
is there something more you are trying to do ?
Regards
Marshall Eubanks
> John Beckmeyer
> beck@pacbell.net
>
