[ppml] IPv6 flawed?

[David Sinn]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1



On Sep 6, 2007, at 11:52 AM, mack wrote:

>> -----Original Message-----
>> From: briand at ca.afilias.info [mailto:briand at ca.afilias.info]
>> Sent: Thursday, September 06, 2007 12:52 PM
>> To: mack
>> Cc: ppml at arin.net
>> Subject: Re: [ppml] IPv6 flawed?
>>
>> "Mack" wrote:
>>> In the most relevant products from Cisco (sup32 and sup720) the IPv6
>> is
>>> stored in a TCAM with a different architecture from the IPv4 TCAM.
>>> Specifically the Sup32 and Sup720-3B both have 256000 IPv4 slots and
>>> 128000 IPv6 slots.
>>
>> Wrong.
>>
>> A closer inspection of the literature, you will find, says *or*, not
>> *and*.
>>
>> Meaning, 256k IPv4 *only*, or 128k IPv6 *only*.
>>
>> And the default set-up is a mix of 192k IPv4 and 32k combined
>> IPv6-and-multicast.
>
> I stand corrected. Is there a command to modify this?
> I was unable to find a documented command to do this.
> The Sup32 and Sup720-3B are basically dead as far as the DFZ are  
> concerned.
> They will continue to work at the edge where routes can be pruned  
> but not at the core.
> If there is not easy way to modify the FIB ratios then these units  
> can't take a full
> set of routes now.  And have not been able to for a substantial  
> amount of time.

Modifying the TCAM division (though potentially more apropos for  
cisco-nsp) can be done by:

(config)#mls cef max ?
   ip            number of ip routes
   ip-multicast  number of multicast routes
   ipv6          number of ipv6 routes
   mpls          number of MPLS labels

Reboot is required and I believe 1000 is the minimum for any one  
category.  Just to make things even more interesting if you modify  
one of these you no longer get the shared nature resource of the  
default config as is shown in:

#sho mls cef max
FIB TCAM maximum routes :
=======================
Current :-
- -------
IPv4 + MPLS         - 192k (default)
IPv6 + IP Multicast - 32k (default)

David

>
>>
>>> An argument can be made that it will be cheaper for some
>> organizations to
>>> switch to IPv6 shortly due to the impending limit of the FIB TCAM on
>> their
>>> routers.
>>
>> Organizations at the edge, maybe, who don't do content or services.
>> Who can deploy IPv6 to IPv4 NAT gateways at their upstream edges.
>>
>> Not content players, and not DFZ players, for sure.
>
> The DFZ core has already upgraded (at least the networks I deal with).
> 256K was insufficient for the routing table plus internal routes  
> months ago.
>
>>
>>> In any case the most relevant Cisco products are going to run out of
>> IPv4
>>> TCAM space before IPv6 even makes a dent.
>>
>> ... which is all the *more* reason why paying close attention to IPv6
>> allocation, aggregation, and usage, is so important.
>>
>> Brian
>
>
> If the TCAM distribution is not modifiable this is not as relevant.
> Mack
> _______________________________________________
> PPML
> You are receiving this message because you are subscribed to the  
> ARIN Public Policy
> Mailing List (PPML at arin.net).
> Unsubscribe or manage your mailing list subscription at:
> http://lists.arin.net/mailman/listinfo/ppml Please contact the ARIN  
> Member Services
> Help Desk at info at arin.net if you experience any issues.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iD8DBQFG4FyaLa9jIE3ZamMRApPlAKCRhmg9G6wzegHPM/RDNHzOiSIeQgCfWmff
8GQB8qMz/VUpb7Wc4M+tcy8=
=PzLw
-----END PGP SIGNATURE-----