Re: Verizon DSL moving to CGN

["Rajiv Asati (rajiva)" <rajiva () cisco com>]

Jack,

I am assuming that you meant MAP, when you wrote MAPS.

> The larger issue I think with MAP is CPE support requirements. There are
> ISP layouts that use bridging instead of CPE routers (which was a long
> term design to support IPv6 without CPE replacements years later). CGN
> will handle the IPv4 issues in this setup just fine. Then there are

I agree. Good point, btw.

This is the classical ISP deployment model, in which the ISP would usually
provide the layer2 modem, and let the customer get the retail CPE.

> those who have already deployed IPv6 capable CPEs with PPP or DHCP in a
> router configuration which does not have MAP support. Given the variety
> of CPE vendors that end up getting deployed over a longer period of
> time, it is easier and more cost effective to deploy CGN than try and
> replace all the CPEs.

Seemingly so, until we start adding up the cost of

        - Logging infrastructure (setup & mtc)
        - Static NAT & Port forwarding (gaming, camera, etc.)
        - CGN redundancy & load-sharing
        - design complexity (to maintain symmetry)
        - ..

> Given US$35/CPE, cost for replacements (not including deployment costs)
> for 20k users is US$700k. CGN gear suddenly doesn't seem so costly.

Let's throw some numbers of the above costs and then we can do the
apple-to-apple comparison. Else, you are right that CGN cost could be a
lot less.


Cheers,
Rajiv

-----Original Message-----
From: Jack Bates <jbates () brightok net>
Date: Monday, April 8, 2013 10:23 AM
To: Tore Anderson <tore () fud no>
Cc: nanog list <nanog () nanog org>
Subject: Re: Verizon DSL moving to CGN

> On 4/8/2013 7:20 AM, Tore Anderson wrote:
> > BTW. It is AIUI quite possible with MAP to provision a "whole" IPv4
> > address or even a prefix to the subscriber, thus also taking away the
> > need for [srcport-restricted] NAPT44 in the CPE.
>
> The problem is NAPT44 in the CPE isn't enough. We are reaching the point
> that 1 IPv4 Address per customer won't accommodate user bases.
>
> The larger issue I think with MAP is CPE support requirements. There are
> ISP layouts that use bridging instead of CPE routers (which was a long
> term design to support IPv6 without CPE replacements years later). CGN
> will handle the IPv4 issues in this setup just fine. Then there are
> those who have already deployed IPv6 capable CPEs with PPP or DHCP in a
> router configuration which does not have MAP support. Given the variety
> of CPE vendors that end up getting deployed over a longer period of
> time, it is easier and more cost effective to deploy CGN than try and
> replace all the CPEs.
>
> Given US$35/CPE, cost for replacements (not including deployment costs)
> for 20k users is US$700k. CGN gear suddenly doesn't seem so costly.
>
> The only way I see it justifiable is if you haven't had IPv6 deployment
> in mind yet and you are having to replace every CPE for IPv6 support
> anyways, you might go with a MAPS/IPv6 aware CPE which the customer pays
> for if they wish IPv6 connectivity(or during whatever slow trickle
> replacement methods you utilize). While waiting for the slow rollout,
> CGN would be an interim cost that would be acceptable. I'm not sure
> there is a reason for MAPS if you've already deployed CGN, though.
>
> I am sure Verizon did a lot of cost analysis.
>
> Jack