[c-nsp] Cisco 10G gear

Gert Doering gert at greenie.muc.de
Sat Jan 17 05:28:40 EST 2015


Hi,

On Sat, Jan 17, 2015 at 10:53:01AM +0200, Chris Knipe wrote:
> The SUP720-3B is perhaps also an option, but if I have to start
> looking at the 720-3BXL then it's becoming very expensive, yet again.

The only difference between -3B and -3BXL is "more routing table memory"
(256k FIB slots vs. 1M FIB slots).  It will not have more pps or bps.

> Do I then also understand correctly that in the case of a SUP720-3B I
> need to purchase a WS-X6708-10G-3C and in the case of a SUP720-3BXL we
> are talking about a WS-X6708-10G-3CXL (there are no 4-port line card
> available with a DFC daughter card), or are these special versions
> with the daughter cards only required to lower the impact of the
> contention on the 8-port cards?

... and I'm fairly sure there is a DFC option for the 6704...

http://www.cisco.com/c/en/us/support/docs/switches/catalyst-6500-series-switches/107258-C6K-PFC-DFC-CFC.html

------------- quote -----------
Q. "What are the benefits of a DFC"

Consider a 6704 module; the 4x 10G ports have the ability to receive up to 60 Mpps of traffic (4x 14.88 Mpps, which is the linerate for 10GE at 64B frames). This is 2 times the amount of traffic that the centralized forwarding engine can handle, and this only account for one slot in the system. More modules in the system can also contribute to this oversubscription.

The addition of a DFC3 to the 6704 module increases the forwarding performance of that module to 48 Mpps; this is just for that slot, so the forwarding performance of the DFC3 is dedicated to the module on which it resides, i.e., not shares. The overall system over-subscription rate is greatly reduced when you add DFCs. The same principles applies to the 65xx modules, which also have optional DFC support, albeit at lower performance levels.
------------- quote -----------


Besides, 15Mpps is quite a bit of traffic - look at your average packet
sizes, and just do the math.  I just looked at one of our customer
ports, and they send in 1.4Gbit/s with 0.3Mpps - so 15Mpps at that
packet size would scale to about 60Gbit/s.

If you do worst-case calculations with 64byte packets, of course 15Mpps
won't get you far - but that's usually not a realistic scenario.  So plan
with realistic traffic value, and add some headroom for the unexpected
(and if you grow faster, just add a DFC later on).

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 291 bytes
Desc: not available
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20150117/ba9eec48/attachment.sig>


More information about the cisco-nsp mailing list