[c-nsp] x6148 vs. x6548

[Geoffrey Pendery]

Excellent information, thanks for that.  In my case, we're looking at
using the 6516A cards with Sup32's, so instead of the 8G fabric we're
looking at sharing the 32G bus.  In theory, if this was the only line
card in the whole chassis (unlikely, and it's not, but just for the
sake of argument) would this actually yield no oversubscription?  Or
even with two 6516A cards?  Or is there an additional bottleneck I'm
missing?


-Geoff


On Thu, Jul 16, 2009 at 5:22 PM, Tim Stevenson<tstevens at cisco.com> wrote:
> Mike, please see inline below:
>
> At 10:45 AM 7/16/2009, M Callahan opined:
>>
>> We have several 3560G switches deployed as access switches in a small data
>> center environment.  With the increased use of Gig at the access layer, the
>> notion of using EtherChannel to get increased uplink bandwidth back to a
>> WS-X6548-GE is something we've been discussing recently too.  As has been
>> discussed, this won't work the way we'd like it to due to the over
>> subscription design of the X6548. That said, would an EtherChannel setup
>> between the SFP ports on the 3560G and a X6516-GBIC card in the 6509 have
>> any similar type of limitation, or would it work to acheive the increased
>> uplink bandwidth?
>
> 6516 is basically 16 1G ports feeding into a single 8G fabric channel, so
> 2:1 oversubscribed if all ports are pumping. But it's not port level
> oversubscription, so you could definitely use this card & it is very
> typically used in this manner.
>
> You just need to engineer it such that enough uplink b/w and agg layer
> capacity exists to support the access layer, based on the environment.
> Oversubscription ratios are very application/environment specific, anywhere
> from 1:1 all the way to 20:1 or more, so some baselining etc might be in
> order.
>
>>
>>
>> Also, is the only difference between the WS-X6516-GBIC and the
>> WS-X6516A-GBIC the per port buffer size?
>
> Of the top of my head, that, and the replacement of Titan/Medusa chips with
> the Hyperion ASIC for multicast replication & fabric interface. Hyperion
> permits the use of egress replication with the A version of the card.
>
> HTH,
> Tim
>
>
>
>
>> Thanks,
>>
>> Mike
>
>
>
>
> Tim Stevenson, tstevens at cisco.com
> Routing & Switching CCIE #5561
> Technical Marketing Engineer, Cisco Nexus 7000
> Cisco - http://www.cisco.com
> IP Phone: 408-526-6759
> ********************************************************
> The contents of this message may be *Cisco Confidential*
> and are intended for the specified recipients only.
>
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>