[c-nsp] Is the 6704 really as terrible as everyone says?

[Chris Evans]

We are weary of enabling qos on existing server farms based off this
reason.... what will it break because of queue limitations...
On Dec 2, 2010 11:44 AM, "Phil Mayers" <p.mayers at imperial.ac.uk> wrote:
> On 02/12/10 16:02, Nick Hilliard wrote:
>> If you're having packet loss due to shallow buffers on any other line
>> cards, make sure you have "mls qos" disabled. By default, this will carve
>> up the buffers on all line cards in a way that would make you weep.
>
> Or presumably reconfigure the queueing?
>
> ! untested!!!
> wrr-queue queue-limit 100 0 0 0 0 0 0
> no wrr-queue random-detect 1
> wrr-queue cos-map 1 1 0 1 2 3 4 5 6 7
>
> ...on a 1p7q8t port like a 6704. Would be interested if other people
> have tried this successfully? We don't bother, but I do know of people
> who've had this problem (dropping packets despite being at <10gbit/sec)
>
> To be fair the PFC3 QoS guide explicitly says:
>
> """Do not enable PFC QoS globally and leave all other PFC QoS
> configuration at default values...."""
>
> ...and goes on to describe the problem.
> _______________________________________________
> 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/