6500 question

Unanswered Question
Jan 18th, 2008

Have a sup720-3B and cards without DFC (WS-X67xx).

Sup720-3B'PFC3 provides 30Mpps = 15Gbps

1.- Sup720 comes with a switching Matrix that provides 720Gbps but I will only be using 15Gbps because I depend of sup720 PFC3, right?

2.- For this 6500, how much switching throughput I will have for every card? Will be 15Gbps divided by the number of cards?

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
plumbis000 Fri, 01/18/2008 - 14:14

I think you are looking at the 6k architecture incorrectly.

The sup720 enables the Switching Fabric, which operates at 256Gbps, so you can have full-duplex backplane communication. The "720" comes from the fact there are 2 20Gbps connections to the fabric for each card, so 20 * 2 = 40 * 9 (9 slots on 6509) = 360 * 2 (full-duplex) = 720.

the 30Mpps is the number of packets the MSFC can process, so all cards with CFC daughter cards will send their packets to the supervisor for all FIB/TCAM lookups, in this case the MSFC can deal with a maximum of 30Mpps. If you install DFC cards (that match the MSFC or better) the FIB and TCAM are downloaded locally to the line card so all forwarding decisions are made on the line card and only packets that are destined for the switch, require fragmentation, or have TTL expired are processed by the MSFC.

Also, since you have 67xx cards, they are fabric enabled so they can communicate with the Sup720 at 256gbps, full-duplex over the fabric (not the Bus).

Hope this makes sense.

aalejo Fri, 01/18/2008 - 14:36

Look at this white paper:

http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/prod_white_paper0900aecd80673385.html

30Mpps are PFC packet forwarding rate,the MSFC can process only 500Kpps.

If you don't have DFC on any Card, the PFC will teld to the Cards where the packert should be forwarded. But, for that to happend the packet must hit first the PFC and thare is where the botleneck comes.

Jon Marshall Fri, 01/18/2008 - 15:14

Hi

1) Yes as far as i know you are right. The PFC supports up to 30Mpps centralised forwarding if you have all fabric-enabled cards. If you have classic cards in the chassis you only get approx 15Mpps on the PFC.

If you want to get any where near the potential throughput of the 6500 with a Sup720 then as you say you need DFC cards in your fabric enabled modules.

See attached link for more details of performance per mix of cards.

http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/prod_qas09186a0080159963.html

2) I don't believe that it is split evenly between the number of cards, i think it is just a total limit which can be used by any card at any time.

Jon

Joseph W. Doherty Fri, 01/18/2008 - 16:18

For 67xx cards w/o DFCs, sup720 provides 30 Mpps for the entire chassis. See http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/product_data_sheet0900aecd801459a7.html and/or (table 2) http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/product_data_sheet0900aecd801459a7.html.

30 Mpps supports about 20 Gbps for 64 byte packets, much more effective bandwidth for larger packets.

So

1) your performance bottleneck will be the pps rate of the sup720 but with a "normal" packet size mix, you should do much better than 15 Gbps.

2) Up to the full capacity of the sup720 but shared across all your cards. (Not clear how capacity is shared if multiple cards want more than sup can currently provide.)

PS:

DFCs provide up to 48 Mpps per card, which is insufficient for wire rate for a 48 gig port card; unlikely you'll hit this. Also, fabric bandwidth to 48 gig port card is oversubscribed, also unlikely you hit this.

Actions

This Discussion