[c-nsp] 10GE card for 7609
Marian Ďurkovič
md at bts.sk
Mon Apr 6 12:11:17 EDT 2009
On Sun, Apr 05, 2009 at 10:58:56PM -0400, Lee wrote:
> On 3/31/09, Richard A Steenbergen <ras at e-gerbil.net> wrote:
> > On Tue, Mar 31, 2009 at 10:01:24AM +0200, Gergely Antal wrote:
> >
> >> I meant that you can not push 40G out of a 6704
> >> even with a dfc attached to it.But you can do it with a 6708
> >> with 1:1 subscription.
> >
> > Worse, some days you can't even get 7G in from a single port on a 6704
> > with the other 3 ports unused. We routinely have problems with ingress
> > interface overruns or egress interface output queue overflows on 6704
> > in that traffic range, and DFC doesn't make any difference.
> >
> > It seems like it is head of line blocking, and TAC's only answer is
> > "those things have no buffers, buy a 6708".
>
> We had a TAC case for input queue drops on a 6704 port - they told us
> it's a hardware limitation. When traffic is switched between two
> ports leading to the same asic you're limited to about 8Gb per port.
???
This is definitely possible with 6704 and SXI:
TenGigabitEthernet2/1 is up, line protocol is up (connected)
Last clearing of "show interface" counters 00:04:57
Input queue: 0/2000/0 (size/max/drops); Total output drops: 0
Queueing strategy: fifo
Output queue: 0/40 (size/max)
30 second input rate 9900042000 bits/sec, 137227 packets/sec
30 second output rate 0 bits/sec, 0 packets/sec
TenGigabitEthernet2/2 is up, line protocol is up (connected)
Last clearing of "show interface" counters 00:04:55
Input queue: 0/2000/0 (size/max/drops); Total output drops: 0
Queueing strategy: fifo
Output queue: 0/40 (size/max)
30 second input rate 0 bits/sec, 0 packets/sec
30 second output rate 9900041000 bits/sec, 137226 packets/sec
No single packet drop in 5 minutes, and the traffic is switched locally
in the ASIC - not going out to fabric:
#show fabric utilization
slot channel speed Ingress % Egress %
2 0 20G 0 0
2 1 20G 0 0
This is on WS-SUP720-3B system with WS-X6704-10GE + WS-F6700-CFC.
With kind regards,
M.
More information about the cisco-nsp
mailing list