[c-nsp] Assigning VLANs on a per-subnet basis

[Asbjorn Hojmark - Lists]

>>> The disadvantage of the 4948's is they can't be stacked.

>> The 4948 is purpose-built for top of rack installations. It
>> makes no sense to stack it.

> There are many places to put a switch. Top of rack is one of
> them.

Yes, the 4948 can be used for other things, but it's build and
positioned for that specific purpose (top of rack).

>>> A stacked system gives you redundant PSU which is nice.

>> A 4948 can have redundant PSUs. They can be AC, DC or a mix.
>> And they're hot-swap.

> Sure. But what about options on ports? Stacked gives you that.

What options? 10/100/1000/10G? The 4948 does all of those.

> Especially the 10G 80K lasers.

You mean 10Gbase-ZR? I think that'll be available on X2 also
sometime soon. All of Cisco is moving from Xenpak to X2,
including the next products for the 6500 (8-port 10G and Sup720
with 2x10G).

And again, for a switch that's meant for sitting at the top of
a server rack, -ZR is hardly all that relevant.

> 4948 switching capacity and actual 10 gig throughput is better,
> I'll agree with that.

The "10G" port on a 3750G sits on an 8 Gbps ASIC... Actually,
the 3750G-16TD is also over-subscribed in terms of bandwidth
because the 26 Gbps sits on a 16 Gbps dual-rotating ring.

> But four ports of ten is never enough..

No? But four ports of '10G' on four 8 Gbps ASICs on a 16 Gbps
dual-rotating ring *is* enough? Four ports of ~4G is fine for
a stack of four switches?

(The 4948-10GE only has to ports of 10G, BTW).

> So at this point, it's a still a trade off. It depends on
> what's best for your situation :-)

Sure. I was merely making the point that the 4948 can't be
stacked, because stacking isn't relevant for what it's built
for.

-A