[ppml] Proposed Policy: IPv6 HD ratio

[Howard, W. Lee]

I tried to do some numbers on this recently, using large telco-ISPs,
cable providers, and dial providers.  

If you assign /48 to end users (whether dial, cable, or T1/T3/x),
and you have, say, 500 end users on a router, that's a /39 per router.  
Say you have several routers (or whatever L3 aggregation device) in a 
POP; let's reserve space for 8, you have /36 per POP.  
I don't know how many POPs you have in a metro area, but let's reserve 
16, which I make to be /32 per metro area.  
If you carve your metro areas into regions (say, Northeast,
Southwest, etc., in the US, or maybe national borders in other
regions, within an AS) you might need another layer of hierarhcy,
up to 16, so now within a rough continent, one large ISP needs a 
/28 in one part of the world.


Because of the additional layers of hierarchy, a large ISP is
assumed to have fewer customer assignments within its allocation
than a smaller one.  The ISP above may only have a couple of 
customers, but address space is reserved at each level of the
hierarchy.  Under some definition, that's less efficient.
That's the argument, anyway.

If you want, you can continue the thought exercise further and
assume some level of competition (8-16 ISPs per region, 4 bits), 
maybe more customers on a single router (2000? 11 bits), and factor
in the 6 continents (3 bits) and find yourself with a /10 used 
before we even start.

Lee


> -----Original Message-----
> From: owner-ppml at arin.net [mailto:owner-ppml at arin.net] On 
> Behalf Of Leo Bicknell
> Sent: Tuesday, May 10, 2005 1:22 PM
> To: ppml at arin.net
> Subject: Re: [ppml] Proposed Policy: IPv6 HD ratio
> 
> 
> In a message written on Tue, May 10, 2005 at 06:55:31AM 
> -1000, Randy Bush wrote:
> > i suspect that some folk may not understand all of the 
> implications. 
> > heck, i probably don't.  but one would seem to be that it makes it 
> > even harder for the smaller folk and not too much harder for the 
> > larger.
> 
> I have wondered why we don't use a flat measurement as we 
> already do, simply relaxed to fit the additional v6 "free use 
> of IP addresses". I have yet to see a good argument that 
> large ISP's have significantly more waste.  Most of them are 
> based on IPv4 notions, where if you allocate a /20 to a POP 
> to aggregate you have to justify all of the /20.  In IPv6 you 
> allocate it a /48 and that's that.  Remembering that we're 
> talking about allocated, and not in use, and from that 
> perspective it seems IPv6 should be more efficient than IPv4, 
> based on the current guidelines.
> 
> I don't know what the number should be, but I'm thinking 50-65%.
> 
> -- 
>        Leo Bicknell - bicknell at ufp.org - CCIE 3440
>         PGP keys at http://www.ufp.org/~bicknell/
> Read TMBG List - tmbg-list-request at tmbg.org, www.tmbg.org
>