[f-nsp] NetIron MLX-4 vs Juniper MX240

George B. georgeb at gmail.com
Tue May 11 19:21:52 EDT 2010


On Tue, May 11, 2010 at 3:56 PM, George B. <georgeb at gmail.com> wrote:
> On Tue, May 11, 2010 at 3:20 PM, Niels Bakker
> <niels=foundry-nsp at bakker.net> wrote:
>> * georgeb at gmail.com (George B.) [Tue 11 May 2010, 09:01 CEST]:
>>>
>>> At the present time, I would not suggest to someone that they
>>> dual-stack even if the hardware can do it.  An IPv6 route consumes 4x
>>> the resources of an IPv4 route just to hold the address so with a
>>
>> [snip]
>>
>> blah, blah, blah.
>>
>> So you recommend people do nothing towards deploying IPv6 right now because
>> in five years their current hardware may not hack it anymore?
>>
>>
>>        -- Niels.
>
> No, I recommend people use separate stacks if you are taking the full tables.

And basically the reason is ... I currently show about 320,000 routes
from one upstream for v4.  The MLX can handle 512,000 v4 routes in
hardware.  That leaves room for about 48,000 ipv6 routes provided the
v4 routing table stops growing today but exactly the opposite is going
to happen.  As what remaining v4 addresses are broken into smaller
pieces, the routing table is going to continue to grow with smaller
prefixes.

As of March 2010 there were 2830 routes in Hurricane Electric's IPv6
table but people have not yet begun to migrate their production
networks in earnest yet.  There are some end user networks (eyeballs)
that have begun to move but not many of the content providers have
shown up in v6 yet with all of their offerings.  Again, pretty much
everyone who is multihomed on v4 will be multihomed on v6 so the
majority of the smaller prefixes are going to show up in both tables.
Once those people begin the migration, a lot of dual-stacked routers
are going to blow up.




More information about the foundry-nsp mailing list