[125556] in North American Network Operators' Group
Re: Rate of growth on IPv6 not fast enough?
daemon@ATHENA.MIT.EDU (Owen DeLong)
Mon Apr 19 17:05:31 2010
From: Owen DeLong <owen@delong.com>
In-Reply-To: <s2o3c3e3fca1004191352h924a3034i35bcd76329c8e847@mail.gmail.com>
Date: Mon, 19 Apr 2010 14:01:03 -0700
To: William Herrin <bill@herrin.us>
Cc: Bryan Fields <Bryan@bryanfields.net>, NANOG list <nanog@nanog.org>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
On Apr 19, 2010, at 1:52 PM, William Herrin wrote:
> On Mon, Apr 19, 2010 at 1:22 PM, Bryan Fields <Bryan@bryanfields.net> =
wrote:
>> On 4/19/2010 10:14, Patrick Giagnocavo wrote:
>>> The eyeball ISPs will find it trivial to NAT should they ever need =
to do
>>> so however, something servers cannot do - you are looking at =
numbers,
>>> not operational considerations.
>>=20
>> LSN is not trivial.
>>=20
>> Here is some unverified calculations I did on the problem of scaling =
nat.
>>=20
>> Right now I'm using 42 translation entries in my nat table. Each =
entry takes
>> up 312 bytes of FIB memory, which is ~12.7 Kib of data in the FIB. =
Mutiply
>> this by 250k users and we have 3,124,237 KiB of FIB entries, or 3.1 =
GiB. This
>> is not running any PtP programs or really hitting the network, I'm =
just
>> browsing the web and typing this email to you.
>=20
> Bryan,
>=20
> Is there some reason we believe we need to scale individual NAT
> systems beyond about 1000 users each in order to have the desired
> impact on address recapture/reuse? Growing towards 7B people in the
> world with, let's say, 4 connected client devices each, grouped 1000
> per NAT box requires 7B * 4 / 1K =3D 28M or 1.7 /8's for the eyeball
> networks before structural overhead.
>=20
> Pushing a carrier NAT process shallow has its own set of complications
> (and certainly isn't trivial) but raw scalability doesn't look like
> one of the problems.
>=20
The hardware cost of supporting LSN is trivial. The =
management/maintenance
costs and the customer experience -> dissatisfaction -> support calls -> =
employee
costs will not be so trivial.=20
These facts make me very glad that my networks will NOT be implementing
LSN in any form.
Owen
