[arin-ppml] Draft Policy ARIN-2012-2: IPv6 Subsequent Allocations Utilization Requirement - revised

[Jason Schiller]

ARIN-2012-2 does a good job of aligning subsequant IPv6 Allocations with 
the new more liberal initial allocation policy.

How ever the clause "Has allocated more than 90% of their serving site 
blocks" essentially eliminates networks that have an addressing plan that 
has  a hierarchical addressing model with multiple levels of aggregation.

It is quite possible to have no available addresses at higher levels in 
the hierarchial addressing model to draw down from.  Essentially you would 
be at 100% allocated at the serving site one level up, but possibly less 
than 90% over all serving sites.

Consider the following example:
- ISP give out /48 to end-sites
- Most heavily loaded edge router has 200 customers on it
   * that is 2^8=256 (or 8 bits) = /40 per edge router
- POP with the most edge routers has 4 edge routers
   * that is 2^2=4 (or 2 bits) = /38 per POP
- Largest region has 63 POPs with edge routers
   * that is 2^6=64 (or 6 bits) = /32 per region
- There are 6 regions
   * (Northeast, Mid-atlantic, Southeast,
      Northwest, Mid-pacafic, Southwest)
   * that is 2^3=8 (or 3 bits) = /30 for the US 
- That rounds to a /28 on the nibble

after some time, a POP with three edge routers ends up having 300 
customers on one edge router. So you draw down another /32 and add it to 
the existing regional /32.  From that new /32 you draw down an additional 
POP /38 and add it to the existing POP /38, and then draw down an 
additional /40 for the edge router that has more than 256 customers.

Do something similar in two other regions and you have no available space 
to draw down from to add another edge router, or build a new hub.

Back in April Heather pointed out this issue when she said "The goal is to 
be able to get a subsequent allocation when, partway through deploying, 
you realize you don't have enough room to subnet."

Since then text supporting this concept has been added and deleted.

Would recomend adding back that text...

- has sufficient actual utilization at their serving sites to continue to 
justify the block size being utilized for all serving sites as specified 
in section 6.5.2, and has an inability to draw down space from a higher 
level in their addressing heirarchy.

___Jason

---------- Forwarded message ----------
Date: Thu, 5 Apr 2012 16:09:35 -0400
From: Heather Schiller <heather.skanks at gmail.com>
To: arin-ppml at arin.net
Subject: Re: [arin-ppml] Draft Policy 2012-2: IPv6 Subsequent Allocations
     Utilization Requirement

I am working on final text and prepping slides for the meeting in a
few weeks.  Does anyone have any feedback on this policy or on the
staff feedback?

The goal is to be able to get a subsequent allocation when, partway
through deploying, you realize you don't have enough room to subnet.
I would really like to hear from people who have encountered this
problem and what their assignment vs subnet utilization was.  I don't
expect the "1* customer or infrastructure allocation or assignment" to
hold -- I would like to hear what reasonable number 1 could be
replaced with.  10? 20? 50?

Should there be some bounds that could be put around what a reasonable
subnet size is, so that you can't game the process by creating large,
empty subnets?  Or is subnet size too unique to the network?

  Thanks,
--Heather



On Mon, 1 Oct 2012, Owen DeLong wrote:

>
> On Oct 1, 2012, at 13:52 , Chris Grundemann <cgrundemann at gmail.com> wrote:
>
>> On Mon, Oct 1, 2012 at 1:49 PM, Owen DeLong <owen at delong.com> wrote:
>>> On Oct 1, 2012, at 09:46 , Chris Grundemann <cgrundemann at gmail.com> wrote:
>>>> On Mon, Oct 1, 2012 at 10:19 AM, Owen DeLong <owen at delong.com> wrote:
>> [...]
>>>>> It means that at least one serving site has enough end-site assignments to justify the size of the serving site blocks.
>>>>
>>>> Sure, but how many is that?
>>>
>>> As specified in the original policy... enough such that the next lower nibble boundary would not leave a 25% minfree.
>>
>> OK, so as an example (to ensure we are on the same page); let's say I
>> split MyISP into 11 regions (11 serving sites) and my largest serving
>> site has 200 subscribers today.
>>
>> So under current initial allocation policy:
>> Since 200 > 192 (75% of 256), I jump from a /40 for each serving site to a /36.
>> Since 11 < 12 (75% of 16), I only bump one nibble and receive a /32 from ARIN.
>>
>> Now, let's say I decide (or hire a new architect/engineer who decides)
>> that I screwed up and really need 15 serving sites (regions) in my
>> network. I split the network up but leave the largest serving site
>> intact (no change = 200 subs).
>>
>> Under this new/proposed criteria:
>> Since 15 > 14.4 (90% of 16), MyISP gets the next largest nibble from
>> ARIN, a /28.
>> Since my largest site has 200 subs, I have a maximum of 3,000 subs
>> total but I now have over 1 million /48s.
>> That puts me at less than 0.3% utilization of /48s.
>>
>> Is that about right?
>> ~Chris
>>
>
> Yep... That's about right.
>
> If you see a problem with this, then, let me know what it is.
>
> Owen
>
> _______________________________________________
> PPML
> You are receiving this message because you are subscribed to
> the ARIN Public Policy Mailing List (ARIN-PPML at arin.net).
> Unsubscribe or manage your mailing list subscription at:
> http://lists.arin.net/mailman/listinfo/arin-ppml
> Please contact info at arin.net if you experience any issues.
>