[v6ops] Re: Dynamic addresses

[Daryll Swer <[e-mail]>]

David

I don't know how Tier 1 and Tier 2 networks design their network
architecture, but in my experience with non-Tier 1 and Tier 2 SP networks,
we have no problems guaranteeing permanently static IPs even if a BNG dies.
VRRP/Stateful HA does its job. If both BNGs die (unlikely, what's the point
of HA if both dies at the same time?), usually, in my experience, the SP's
Tier 1 engineers, will get out of their house at 2am and swap the box, it's
up and running by 5am. NO IP changes.

If somehow we are forced to re-route traffic to a completely different site
over the MPLS backbone, it's not an issue to also re-route the aggregated
/44s or /40s (whatever it may be) from dead BNG to new BNG site with
automation (Ansible? Custom Python? Etc).

In short, again, we don't charge for *static* IPv6, full stop. But yes, if
a customer wants *additional *IPv6 (> /56 or > /48), then the additional
prefix is charge-able. But at that point, many customers would just do PIA
anyway.

*--*
Best Regards
Daryll Swer
Website: daryllswer.com
<https://mailtrack.io/l/339ec4c2393571ab643dd0d7d612f76ea7057139?url=https%3A%2F%2Fwww.daryllswer.com&u=2153471&signature=832f920953617652>


On Fri, 16 Aug 2024 at 20:16, David Farmer <farmer=[e-mail]>
wrote:

>
> On Fri, Aug 16, 2024 at 05:45 <[e-mail]> wrote:
>
>> Hi,
>>
>> *Von:* Daryll Swer <contact=[e-mail]>
>> *Gesendet:* Montag, 12. August 2024 06:07
>> *An:* Brian Candler <[e-mail]>
>> *Cc:* The Multach's <[e-mail]>; [e-mail]
>> *Betreff:* [v6ops] Re: Dynamic addresses
>>
>> Getting a new IPv4/IPv6 allocation on session disconnect and reconnect
>> is a matter of network design.  If the network design is that aggregate
>> address pools are routed to BRASes, and the end user's address is
>> allocated by the BRAS from its pool, then when you reconnect to a
>> different BRAS you'll get a different address. So be it.
>>
>> > Yeah, they/we typically route an aggregate pool to the BRAS/BNG. But
>> also,
>>
>> > they are configured in HA mode with VRRP etc and the pools do not
>> change.
>> > If they did change, we now have the problem with connectivity stability
>> again,
>>
>> > and this brings up the old conversation about DHCPv6 HA (vendors solved
>> it,
>>
>> > to my knowledge, using proprietary software).
>>
>> That’s what we are doing as well. We aggregate on the BNG and assign
>> addresses
>> (IPv4 and IPv6) from the address pools of the BNG. For residential users
>> those addresses are dynamic.
>>
>>
>>
>> For business customers requiring a static address/prefix we backhaul them
>> to a different set of BNGs providing static addresses. Those BNGs have
>> their own pools for static assignments and allow to use the same, static
>> address
>> even in case users are moving within our network (e.g., from one city to
>> another city).
>>
>
> On one level, this seems reasonable; static customers pay more for the
> extra backhaul service to different BNGs with static pools. I understand
> that, and it appears acceptable to me.
>
> Regular customer addresses can change, and that's acceptable on one level.
> But how often do they change, and what triggers these changes? Is it simply
> due to rebooting and acquiring new addresses with each reboot? Or does it
> occur during customer rebalancing events, where the BNGs for some customers
> are changed, probably in an overnight change window? I'd like to know the
> frequency.
>
> While you can change their addresses, even regular customers deserve care
> and stability. Their addresses shouldn’t be altered haphazardly, too
> frequently, or unnecessarily.
>
> With my ISP, when I reboot, I usually get the same address. However, my
> address changes occasionally, I estimate once or twice a year. It is
> primarily stable but changes infrequently, which seems reasonable to me.
>
> There likely needs to be a service distinction between the statically
> addressed business class and the dynamically addressed consumer class
> customers. But there is a difference between a dynamic where addresses are
> changed occasionally, like once a month or less, and a crazy dynamic where
> addresses are changed haphazardly on every reboot or an overly frequent
> basis, like every few hours or daily.
>
> Thanks.
>