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

[George B.]

"they seriously lack memory when it comes to dualstacking ipv6 and ipv4"

I believe that is going to be a major challenge for all the vendors
and the first one to seriously load up their gear with a lot of CAM
and not charge out the wazoo for it is going to win bigtime.

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
given amount of resources you are going to be able to manage 1/4 of
the routes.  That isn't so bad these days because the v6 routing table
is fairly small but consider that everyone that is currently
multihomed on v4 will probably be multihomed on v6 and will be in v6
PI space.  And it gets worse because they will probably be in PI space
in both v4 and v6 at the same time.  So it is my opinion that vendors
need to have about 5x more resources on the hardware than they
currently do.  They are still treating v6 as some future issue that
isn't a problem until it is a problem.  The only way around this in
any sustainable fashion is to have separate stacks if you are taking
full routes as the v6 routing table is likely to become as large as
the v4 table (actually has the potential to become even larger when
you consider how many routes smaller than a /32 are going to be
announced.)  We have a /44 block.  ARIN is issuing /48's.  Imagine
every single host on the current internet having its own route but
that route consumes 128 bits instead of 32.  And we are only talking
about /32 prefixes and not the more specifics that will be multihomed.

 So far I haven't seen anyone's gear that is prepared to scale to the
size that dual stacking will require once there is serious migration
to v6.  The reason is that nobody is going to give up their v4 space
and will be using both spaces for likely at least the next 10 years or
more.

On Mon, May 10, 2010 at 11:37 PM, Andreas Larsen <andreas at larsen.pl> wrote:
> We have 16 MLX 8 installed in our network. They are good however they
> seriously lack memory when it comes to dualstacking ipv6 and ipv4 I wouldn't
> today buy a XMR rather to be more future proof.
>
> Regards Andreas
>
> On Tue, May 11, 2010 at 7:31 AM, Richard A Steenbergen <ras at e-gerbil.net>
> wrote:
>>
>> On Mon, May 10, 2010 at 09:22:17PM -0700, David Kotlerewsky wrote:
>> > Samit,
>> >
>> > List price in MX80 will be 40K for the model with fixed 48x RJ45 ports
>> > + 4x 10GbE XFP ports. A similarly built MLX-4 config will run at
>> > $79,730 (base chassis with 1 mgmt module, 2 fabric modules, 1 AC
>> > power+fan is $19,245. Then you need to add 1 4x10GbE XFP module at
>> > $27,495 and 2x MLX 20-port 1GbE RJ45 modules at $16,495 each). Now you
>> > may get a very nice discount from Brocade, but Juniper will be game as
>> > well in trying to market the MX80, so you may see some really nice
>> > discounts. Also, the MX80 is only 2U of space, and the MLX-4 is 4U.
>>
>> $40k is just the base, it's another $20k for the full routing license.
>>
>> --
>> Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
>> GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)
>> _______________________________________________
>> foundry-nsp mailing list
>> foundry-nsp at puck.nether.net
>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp at puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>