headers
Steve Bertrand | 2 Apr 2008 03:19
Picon

BCP: Slicing a /32 for an ISP

Quick to the point:

Has anyone done any documentation on how they decided to slice up their 
IPv6 allocation?

I'm pulling hairs deciding on a tactical solution for the long-term in 
regards to how I should 'cut up' my /32.

Can anyone here provide any feedback, positive or negative to the way 
they have done this?

What did you use for core infrastructure? What did you divide off for 
remote PoPs? Did you cut up your /32 into /35's and work from there? etc 
etc.

I can't find any decent documentation for what has worked for an ISP, 
and how 'they' did it, particular to a small ISP. (forgoing an argument 
about CPE assignment, I just want to know how people are assigning and 
delegating routing slots from the core to their edge).

Before I can make a concrete decision (because the size of the space 
still intimidates me), I've been simply using 'my:32::x/64' type 
addresses for testing within a single subnet.

Regards,

Steve
Permalink | Reply | 26 comments |
headers
SM | 3 Apr 2008 23:53

Re: current usage of AAAA implicit MX?

Hello,

I'm copying a message from Tony Hansen who is the shepherd for RFC 2821bis.

At 13:32 03-04-2008, Tony Hansen wrote:

Before we can put the issue of AAAA implicit MX to bed, we need to 
get an answer to the question of how much AAAA implicit MX is already 
being used in the real world.

It's been postulated that the current deployment of this and its use 
is low. However, RFC 3974 (SMTP Operational Experience in Mixed 
IPv4/v6 Environments), was published three years ago, and it mandates 
AAAA implicit MX (as much as an informational document can mandate anything).

So I'm looking for information from people who are currently working 
in the IPv6-based email world.

I'm assuming that a number of email sites over in Asia are already 
IPv6-enabled, and there are probably other enclaves as well. So I'm 
looking for information on actual email deployment and sites.

I'm looking for help answering these questions:

   *	How are existing IPv6-enabled mail sites actually dealing with
	missing MX records?

   *	Are the IPv6-enabled email sites mostly following the
	recommendations found in RFC 3974?
(Continue reading)

Permalink | Reply | 25 comments |
headers
Mark Tinka | 14 Apr 2008 03:48

Re: BCP: Slicing a /32 for an ISP

On Wednesday 02 April 2008, Steve Bertrand wrote:

> Can anyone here provide any feedback, positive or
> negative to the way they have done this?

We did the following (this applies just to our network - 
different strokes for different folks):

* From the allocated /32
* Carve out a /48 per PoP

* From each /48 assigned to each PoP
* Carve out /64's for various services
* From the main PoP's /64, carve out a set of:

		- network-wide Loopback addresses

* From /48's assigned to each PoP
* Carve out /64's for each of the following:
	
		- routing & switching infrastructure
		- production servers & services
		- WAN point-to-point links (customers)

* Drilling deeper into each PoP's assignment:

		- /128 for Loopback addresses
		- /112 used for routing & switching kit
			* border
			* core
(Continue reading)

Permalink | Reply | 17 comments |
headers
Michael Adams | 14 Apr 2008 08:03
Picon
Favicon

Re: BCP: Slicing a /32 for an ISP

Mark Tinka,  14 Apr 2008 9:48:

> * Drilling deeper into each PoP's assignment:
> 
> 		- /128 for Loopback addresses
> 		- /112 used for routing & switching kit
> 			* border
> 			* core
> 			* edge
> 			* peering
> 	
> 		- /112 used for production servers & services
> 		- /126 used for WAN point-to-point links

I'm eager to hear more opinions about this. We are going to use
/64 for all kind of network interfaces including point-to-point 
links and loopback interfaces. 

regards,
Michael

--

-- 
Michael Adams                             Tel: +49 221 2222 657
Network Engineering & Design              Fax: +49 221 2222 7657

NetCologne                                Geschäftsführer
Gesellschaft für Telekommunikation mbH    Werner Hanf
Am Coloneum 9                             Dipl.-Ing. Karl-Heinz Zankel
50829 Köln                                HRB 25580, Amtsgericht Köln
(Continue reading)

Permalink | Reply | 13 comments |
headers
Mohacsi Janos | 14 Apr 2008 09:21
Picon
Favicon

Re: BCP: Slicing a /32 for an ISP


On Mon, 14 Apr 2008, Michael Adams wrote:

> Mark Tinka,  14 Apr 2008 9:48:
>
>> * Drilling deeper into each PoP's assignment:
>>
>> 		- /128 for Loopback addresses
>> 		- /112 used for routing & switching kit
>> 			* border
>> 			* core
>> 			* edge
>> 			* peering
>>
>> 		- /112 used for production servers & services
>> 		- /126 used for WAN point-to-point links
>
> I'm eager to hear more opinions about this. We are going to use
> /64 for all kind of network interfaces including point-to-point
> links and loopback interfaces.

We are using /64 mainly where necessary. For point-to-point link we are 
not uisng anything, since it is not neccesary: routing protocol are using 
link-local for IPv6 routing.

>
> regards,
> Michael
>
> --
(Continue reading)

Permalink | Reply | 2 comments |
headers
Marco d'Itri | 14 Apr 2008 09:16
Picon
Favicon

Re: BCP: Slicing a /32 for an ISP

On Apr 14, Michael Adams <madams <at> netcologne.de> wrote:

> I'm eager to hear more opinions about this. We are going to use
> /64 for all kind of network interfaces including point-to-point
> links and loopback interfaces.
I recommend against this, because if you use longer prefixes then you
can encode useful data and/or a hierarchy in the address.

--

-- 
ciao,
Marco
Permalink | Reply | 1 comment |
headers
Christian Hahn | 14 Apr 2008 09:31
Picon

Re: BCP: Slicing a /32 for an ISP


Steve Bertrand schrieb:
| Quick to the point:
|
| Has anyone done any documentation on how they decided to slice up their
| IPv6 allocation?
|
Hi Steve,

take a look at the following IETF Draft
http://www.ietf.org/internet-drafts/draft-ietf-v6ops-addcon-07.txt
It's on the way to become RFC and includes two case studies for address planing
for Enterprises and ISPs in the appendix. Hope this helps.

best wishes,
Christian

| I'm pulling hairs deciding on a tactical solution for the long-term in
| regards to how I should 'cut up' my /32.
|
| Can anyone here provide any feedback, positive or negative to the way
| they have done this?
|
| What did you use for core infrastructure? What did you divide off for
| remote PoPs? Did you cut up your /32 into /35's and work from there? etc
| etc.
|
| I can't find any decent documentation for what has worked for an ISP,
| and how 'they' did it, particular to a small ISP. (forgoing an argument
| about CPE assignment, I just want to know how people are assigning and
(Continue reading)

Permalink | Reply | 0 comments |
headers
Simon Lockhart | 14 Apr 2008 09:33
Favicon

Re: BCP: Slicing a /32 for an ISP

On Mon Apr 14, 2008 at 09:21:22AM +0200, Mohacsi Janos wrote:
> We are using /64 mainly where necessary. For point-to-point link we are 
> not uisng anything, since it is not neccesary: routing protocol are using 
> link-local for IPv6 routing.

I'm currently debating whether to use public IPs for point-to-point (or
inter-router ethernets). Apart from "documentation" purposes, is there any
good reason not to just use link-local, and then use loopbacks for stuff that
requires a router to use an IPv6 address (e.g. iBGP)?

Simon
--

-- 
Simon Lockhart | * Sun Server Colocation * ADSL * Domain Registration *
   Director    |    * Domain & Web Hosting * Internet Consultancy * 
  Bogons Ltd   | * http://www.bogons.net/  *  Email: info <at> bogons.net  *
Permalink | Reply | 1 comment |
headers
Mohacsi Janos | 14 Apr 2008 09:55
Picon
Favicon

Re: BCP: Slicing a /32 for an ISP


On Mon, 14 Apr 2008, Simon Lockhart wrote:

> On Mon Apr 14, 2008 at 09:21:22AM +0200, Mohacsi Janos wrote:
>> We are using /64 mainly where necessary. For point-to-point link we are
>> not using anything, since it is not necessary: routing protocol are using
>> link-local for IPv6 routing.
>
> I'm currently debating whether to use public IPs for point-to-point (or
> inter-router ethernets). Apart from "documentation" purposes, is there any
> good reason not to just use link-local, and then use loopbacks for stuff that
> requires a router to use an IPv6 address (e.g. iBGP)?

The only issue we see, since we are using in our backbone link-local for 
point-to-point: traceroute. You have to configure the source address ipv6 
udp unreachable to be picked up deterministically e.g. from loopback. 
(ipv6 unnumbered if - Cisco terminology).

>
> Simon
> -- 
> Simon Lockhart | * Sun Server Colocation * ADSL * Domain Registration *
>   Director    |    * Domain & Web Hosting * Internet Consultancy *
>  Bogons Ltd   | * http://www.bogons.net/  *  Email: info <at> bogons.net  *
>
Permalink | Reply | 0 comments |
headers
Mark Tinka | 14 Apr 2008 09:57

Re: BCP: Slicing a /32 for an ISP

On Monday 14 April 2008, Michael Adams wrote:

> I'm eager to hear more opinions about this. We are going
> to use /64 for all kind of network interfaces including
> point-to-point links and loopback interfaces.

We saw no point in using /64 especially for point-to-point 
links and Loopback interfaces.

At my previous employer, our upstream provider decided to 
use a /64 on a point-to-point transit link. It was their 
network, but I could have done it differently - besides the 
2 routers that talk to each other over this circuit for a 
specific purpose, what other network elements are you 
possibly going to have in between that justifies allowing 
for 18,446,744,073,709,552,000 addresses, as opposed to 
just 2.

We, more or less, used the same argument for our LAN 
segments. If our core network comprises a handful of 
routers and switches, I'm sure even 65,536 addresses for a 
few hundred nodes should be sufficient.

Cheers,

Mark.
Permalink | Reply | 2 comments |

Gmane