[c-nsp] BGP: 7206vxr-G1 vs. 6503 sup2

[Simon Leinen]

Chris Cappuccio writes:
> Christopher McCrory [chrismcc at pricegrabber.com] wrote:
>> Hello...
>> 
>> I'm adding more horsepower in my BGP core.  I am currently running
>> multiple 7206VXR-300 (12.0S) with BGP/OSPF/IPv4 (for a content
>> provider network, not ISP).  My transit connections are all same
>> floor FastE/GigE cross connects with a backup DS3.
>> 
>> Q:  any advantages to using a 6503/sup2/msfc2 instead of a
>> 7206VXR-NPE-G1 ?

> Yeah, on the 6500 it's all hardware forwarding and therefore much,
> much faster

Right, 30 Mpps rather than about 1 Mpps for IPv4 traffic.

> But you also get less options in software and hardware, but at the
> core it shouldn't matter much....

Yes.  If you're interested in hardware IPv6 (including multicast and
IPv6-in-IPv4 tunneling) or MPLS forwarding (including EoMPLS encap),
look at the Sup32 (cheap) or Sup720 (includes fast switching fabric).

Since you say "BGP core", you should also take the number of routes
you want to carry into account.  The PFC2 (on the Sup2) supports ~250k
routing table entries in hardware (but drops to half that if you
enable unicast RPF).  PFC3 (A or B) for Sup32/Sup720 supports the same
number, but enabling uRPF doesn't reduce that.  The PFC3B-XL for
Sup32/Sup720 supports about 800k IPv4 routes in hardware as far as I
recall.

Since the NPE-G1 does forwarding in software, number of routes should
only be limited by amount of RAM.
-- 
Simon.