[c-nsp] OSPF cost calculation on ASR9K

[Saku Ytti]

On (2014-01-08 13:55 +0200), Mark Tinka wrote:

> > Cannot be, I remember we got few configs changed
> > interface cost. Which version u r running.
> 
> I'm into manually setting cost a la IS-IS, anyway.
> 
> If I were running OSPF, I'd probably do the same, as my idea 
> of an inherent reference bandwidth is different from what 
> vendors might think.

Absolutely. STP cost and IGP metric should be role-based not bandwidth based.

You (should) know where you want traffic to flow, you have logical map of your
network in your head and you know for PE1 to to PE2, it should go
PE1-P1-P2-PE2, not PE1-PE3-PE4-PE2, even if that path is available.  Or access
switch should prefer to go to access-core-access, not access-access-access.

If you don't have bandwidth, you either to cost/metric hacks, or you add
bandwidth.
It makes no sense to shift new traffic to interface, just because you upgraded
its capacity to handle organic growth of existing traffic.

-- 
  ++ytti