[c-nsp] Making SUP720 cope better under BGP load

Nick Hilliard nick at foobar.org
Sun Dec 9 09:02:22 EST 2012


On 09/12/2012 12:08, Chip Gwyn wrote:
> The other issue at hand is going to be route table size.  If I recall
> correctly TCAM on the 2T is the same as the 720.

more correctly, it comes in two primary variants, the pfc4 (256k ipv4
entries) and the pfc4xl (1m ipv4 entries).  These numbers correspond
exactly to the capacity of the pfc3 / pfc3-xl models, but that's not very
interesting because it doesn't describe any of the massive differences
between the two sups.

There's a full set of details at this URL:

> http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps708/white_paper_c11-676346.html

but very briefly, the sup2t has the following advantages over the sup720:

- 9 years of development
- huge increase in cpu power on the MSFC5
- forwarding capacity is much greater
- netflow actually works
- support vpls on lan cards
- enhanced ACL support
- ipv6 urpf
- supports CMP
- support ingress policing
- much better CoPP support
- many more things supported in hardware that weren't supported properly on
pfc3

In short, the sup2t is a vast improvement on the sup720.  Getting back to
the OP, a sup720 is not really suitable as a DFZ router these days, and I
would not connect one up to an IXP the size of LINX because of the RP
speed's impact on BGP convergence times and due to problems with the IOS
BGP scheduler which causes keepalive starvation and session flaps.  A sup2t
should work fine in the same situation, but I would really recommend a
better BGP implementation for this (e.g. XR).

Nick

> Sent from my mobile device, please excuse any typos.
> 
> On Dec 9, 2012, at 5:45 AM, Tóth András <diosbejgli at gmail.com> wrote:
> 
>> Adrian,
>>
>> Inband channel (link to CPU) is a 1GE full-duplex link in both Sup720 and
>> Sup2T.
>>
>> Best regards,
>> Andras
>>
>>
>> On Sun, Dec 9, 2012 at 10:25 AM, Adrian Minta <adrian.minta at gmail.com>wrote:
>>
>>> On 12/09/12 07:10, Andrew Jones wrote:
>>>
>>>> Sup720 cpu is around 600mhz if i remember correctly, whilst sup2t is 1.5
>>>> ghz dual core, so one would sup2t would handle this much better. Also,
>>>> sup2t has much better CoPP capability with built in default config
>>>> templates, ready for you to tune if needed.
>>>>
>>>>
>>>> The CPU network interface is still 10mbps half duplex ?
>>>
>>> --
>>> Best regards,
>>> Adrian Minta
>>>
>>>
>>>
>>> ______________________________**_________________
>>> cisco-nsp mailing list  cisco-nsp at puck.nether.net
>>> https://puck.nether.net/**mailman/listinfo/cisco-nsp<https://puck.nether.net/mailman/listinfo/cisco-nsp>
>>> archive at http://puck.nether.net/**pipermail/cisco-nsp/<http://puck.nether.net/pipermail/cisco-nsp/>
>>>
>> _______________________________________________
>> 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/
> 
> _______________________________________________
> 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/
> 



More information about the cisco-nsp mailing list