[c-nsp] feasible number of routing adjacencies on a C76K

[Tim Franklin]

Hi Zoltan,

>    Given a network providing L3 VPN service, with C7600s 
> running with 12.2SX as PE routers. Due to redundacy 
> requirements dynamic PE-CE routing has to be deployed
> in each VPN but since average traffic/load per VPN is rather 
> low/small customer is tempted to terminate over hundred VPNs 
> per PE (potentially over 300 hundred)
>  
> For a reason customer is fallen in love with OSPF and since 
> 12.2(18)SXE the maximum number of OSPF processes is not 
> hard-limited anymore but:
>  
> I want to push them to RIPv2 on PE-CE:
>  - as the topology is very-very-very simple; just 2 CE RTRs 
> per VPN (internal C topology is deemed to be simple as well)
>  - as is it much simpler than OSPF
>      --->  so the practical routing ADJ limit is higher in case of RIP
>  - as it won't require a process per VPN,  just a routing context
>  - as it won't bind them to one and only sw release (as of 
> now) namely 12.2(18)SXE (which, I don't belive, is the sw 
> train we shall use today)

This sounds very sensible - we've settled on RIP or BGP only (or statics, of
course) as PE-CE routing protocol for exactly the same reason of routing
context vs process per VRF.

Quick browse of configs shows our 7600s running typically 150-200
address-families per box (mixture of RIP and BGP), but I'm seeing some 300s
and 400s in there without issue.  Routes per VRF is anywhere between five
and five hundred, although typically in the tens.

Sounds to me like you shouldn't have any issues with scaling to the numbers
you're talking about.

Regards,
Tim.

-- 
____________   Tim Franklin                 e: tim at colt.net 
\C/\O/\L/\T/   Product Engineering Manager  w: www.colt.net 
 V  V  V  V    Managed Data Services        t: +44 20 7863 5714 
Data | Voice | Managed Services             f: +44 20 7863 5876