[c-nsp] Broadband Aggregation/Termination

[Aaron Gould]

Careful lease time adjustments make this a non-issue with dhcp (referring to Andrew's comment below).  I've worked this ISP for 9 years and have churned ipv4 space all over the place month after month after month and there are simple, elegant, non-customer-impacting ways to change customer subnets.  Can you cause an outage... yes... can you throw/churn subnets and not cause an outage... yes.  You gotta know what you are doing of course.  It's not hard.  Again, it's simple, back the lease times down, lower and lower until you have dhcp clients requesting and hitting half-life fairly quickly so during the throw, you have the ability to throw back pretty quickly and get the customer dhcp clients to re-request in short order.  After things are settled down and you are satisfied, you move the lease times back up to 4 hours, days, weeks, etc.  Ummm, don't run heavy dhcp logging in your servers when you crank lease times down or you could bury your server... again, just know what you are doing :)  ...can you tell I've learned this the hard way :)

IPv6 has been tested in my dsl, ftth and cable modem communities.  Dual stacking seems to be the way to go from my limited v6 perspective.  Dual stacking on my MPLS PE edge routers works quite nicely over the exsiting mpls l3vpn's... known as 6VPE.

I'm moving forward with more v6 testing and my target communities will be the aforementioned cgnat'd dsl communities.  I've already tested this dual stacking of rfc1918 ipv4 with publicly routable ipv6 and works nicely... I have ipv6 flow unchanged out to the internet and ipv4 flows via dual cgnat nodes (juniper mx104 w/ms-mic-16g)

My boss had me do an ipv6 impact study on my actual cpe for ftth, dsl and cable modem a few years ago and noted a lot of info... one fine point comes to mind now that I had to change something in a profile on the occam (calix) B6 series dslam to allow v6 RA's (router advertisements) to flow correctly... I recall that basic first-hop routing wasn't working until I did that.

- Aaron

-----Original Message-----
From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Andrew Gambell
Sent: Wednesday, April 19, 2017 12:04 PM
To: CBL <alandaluz at gmail.com>; Cisco <cisco-nsp at puck.nether.net>
Subject: Re: [c-nsp] Broadband Aggregation/Termination

Hi Cassidy,

....
Session maintenance/timeout is a pain in the arse with DHCP with the only real option to rely on the DHCP lease, if you don’t use a low lease time this makes router maintenance or moving subscribers between devices a long outage for customers while waiting for CPE to request a new lease. Whereas with PPPoE we have keep alives at typically 30-60 seconds which means session tear down and come back quickly.
.....
--Andrew Gambell