Re: Definitive Guide to IPv6 adoption

[Owen DeLong <owen () delong com>]

On Oct 18, 2010, at 9:59 AM, Jack Bates wrote:

> On 10/18/2010 11:47 AM, Randy Carpenter wrote:
> > Unfortunately, it is not as easy as that in practice.
> >
> > I recently worked with a customer that has ~60,000 customers
> > currently. We tried to get a larger block, but were denied. ARIN said
> > they would only issue a /32, unless immediate usage could be shown
> > that required more than that. Their guidelines also state /56 for
> > end-users. I am a big proponent of nibble boundaries, too. I think if
> > you are too big to use only a /32, you should get a /28, /24, and so
> > forth. It would make routing so much nicer to deal with.  /31 and
> > such is just nasty.
>
> ARIN does reservations (unsure at what length, but at least down to /31). If you were to fill the /32 quickly, you 
> could easily request the next block. To my knowledge, they've only handed out 1 or 2 networks shorter than /32.
Not any more...

ARIN now uses allocation by bisection.

> Correct me if I'm wrong, but isn't 60,000 customers at /56 2^24 assignments from a /32? Seems plenty. Even at /48 
> assignments, you'd get 65,536 assignments. So how can you justify more than a /32?
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.

60,000 customers may well be more than 65,536 end sites. Also, you need to leave room for numbering infrastructure, 
sizing POPs to prefixes, etc.

It's much more complex than just number of customers = number of /48s.

Unfortunately, current policy doesn't recognize that other than HD ratio. However, 60,000 customers each with a /48 
would far exceed the .94
HD ratio requirement for larger than a /32. IIRC, under current policy it would justify a /30 or possibly a /29.

Owen