[c-nsp] ipv6

[Lobo]

Here is what we sort of followed:

  * Get your IPv6 block from your RIR; typically a /32 but we were able
    to negotiate a /28
  * Come up with a good IPv6 address plan; spend some good time on this
  * Enable IPv6 connectivity to your upstream and public peering
    connections along with BGP
  * Enable dual stack on your backbone and edge devices (routers) and
    decide which IGP you will use (OSPFv3 or ISIS)

At that point you should be able to ping via IPv6 from any backbone/edge 
router to another.

With regards to the IP addressing scheme, we followed a breakdown of 
assigning /34(s) to each market we're in (about 16).  Those market /34s 
were then broken up into /48s to be handed out to our customers who in 
turn will be able to break them down into /56s or /64s.  We chose /48s 
because we deal with enterprise customers (no residential) and the 
general rule (as per ARIN) is to give them a /48 block.  Even with this 
rather "wasteful" allocation, we still will not come anywhere close to 
chewing up our space in the coming decades.  :)

With regards to servers and IT related things, we've struggled in that 
dept but we've managed to convince IT of the importance of at least 
having a public DNS server with IPv6 access (dual stacked) for our beta 
customers.

Still have lots to learn and we're really hoping to avoid any instances 
of CGN if at all possible.

Jose



On 6/5/2013 5:00 PM, Jay Ford wrote:
> On Wed, 5 Jun 2013, Aaron wrote:
>> There seems to be so many ways to do ipv6..(I'm not clear on what to 
>> use and
>> why to use it and when to use it..)
>>
>> I work for an isp of about 30,000 customers.
>> -          Ftth
>> -          Dsl
>> -          Cable modem
>>
>> What is the best way to migrate my customers to IPv6 with zero impact
>> (meaning, all internet and services are still reachable when done 
>> moving to
>> ipv6) ?
>>
>> I don't need configs and technical details, just a technique or 
>> technology
>> answer for now will suffice to get my moving in a research direction.
>
> The best general approach is to add native IPv6 alongside IPv4 & plan 
> to run
> that way for several years.  Most clients will now do the right thing 
> when
> presented with working IPv4 & IPv6 connectivity, so your job is to add
> working native IPv6.  You should try very hard to avoid NATed IPv6, so if
> your IPv4 is NATed now you'll have to keep that difference in mind.
>
> This sequence worked well for us:
>    o  devise an IPv6 address plan
>    o  get IPv6 address space from ARIN
>    o  establish native IPv6 connectivity upstream
>    o  enable IPv6 in your routed net
>    o  enable IPv6 to net-related services (DNS, NTP, syslog, SNMP...)
>    o  enable IPv6 on the net staff desktop net, so they can & in fact 
> have to
>       use it every day
>    o  educate users, at least those who know what IP is
>    o  enable IPv6 to some "early adopter" users
>    o  enable IPv6 to the rest of the users
>
> It really wasn't that hard for us to roll out native IPv6.  It took a 
> while,
> but that's a reason to get started rather than delay.  There are some 
> things
> which might not yet be up to IPv6, mainly things which try to be extra 
> smart
> (firewalls, load balancers, home gateways...) & things which are dumb
> (printers, embedded control devices, home gateways...).  Yes, I know I 
> put
> home gateways in both lists. ;^) Most things in the middle of the device
> spectrum behave pretty well these days.
>
> ________________________________________________________________________
> Jay Ford, Network Engineering Group, Information Technology Services
> University of Iowa, Iowa City, IA 52242
> email: jay-ford at uiowa.edu, phone: 319-335-5555, fax: 319-335-2951
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/