[130959] in North American Network Operators' Group
Re: Definitive Guide to IPv6 adoption
daemon@ATHENA.MIT.EDU (Owen DeLong)
Mon Oct 18 15:09:35 2010
From: Owen DeLong <owen@delong.com>
In-Reply-To: <Pine.LNX.4.61.1010181412280.5148@soloth.lewis.org>
Date: Mon, 18 Oct 2010 12:00:25 -0700
To: Jon Lewis <jlewis@lewis.org>
Cc: nanog@nanog.org
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
On Oct 18, 2010, at 11:18 AM, Jon Lewis wrote:
> On Mon, 18 Oct 2010, Owen DeLong wrote:
>=20
>> The customers should get /48s. The /56 guideline is merely that and =
only for the smallest of sites. It's also subsequently turned out to be =
bad advice.
>=20
> Can you elaborate on why /56 is "bad advice" and if you're saying it =
only for this case or if you're saying assignment of /56 to any =
customers is a bad idea? Dealing with a data center where customer =
machines typically get by today with a /29 of IPv4, is a /56 really not =
enough for their forseeable future?
>=20
I think it's generally a bad idea. /48 is the design architecture for =
IPv6. It allows for significant innovation
in the SOHO arena that we haven't accounted for in some of our current =
thinking.
In a datacenter environment, you might want to actually assign /64s to =
needed subnets, but, in a
situation where you are serving remote end-sites, a /48 per end-site is, =
IMHO, the minimum
size that should be issued.
> I realize our /32 could support more customers than we're likely to =
fit in the data center at /48 per customer, but is that enough of a =
reason to assign 65k /64 subnets to each customer machine?
>=20
Datacenter is a whole different ball of wax. Nothing wrong with giving =
your customers /48s,=20
but, the right size in a datacenter may well depend on a lot of things =
about your business
model, the nature of your customers, etc.
Certainly I would not deny a /48 to any customer that requested one.
Owen
