[c-nsp] BGP ORR - experiences

[Saku Ytti]

We use ORR and add-path 3 (to enable ECMP and to have backup) with
about 40 ORR perspectives and +15M RIB. But this is Junos running on
compute.

You really want compute, not actual router hardware as the CPU gains
from modern compute are very significant.


I expected we'd have some rearing problems doing this, but it's been
basically problem free since we started about 4 years ago, compared to
the significant amount of issues we had due to full-mesh prior to it.
At the time of doing this, we had some of our other vendors say flatly
'no' to the idea of having 40 ORR views with the RIB scale we have.



Juniper limits the BGP window by default to 16kB, which you can raise
to 64kB, but we'd really want window scaling support too, as our
initial convergence is still bound by the TCP window.

On Tue, 14 Oct 2025 at 20:17, Darko P via cisco-nsp
<cisco-nsp at puck.nether.net> wrote:
>
> Hi everyone,
>
> I just wanted to ask about experiences with running BGP ORR, or if anyone
> actually uses it.
>
> I know Cisco faced significant issues with the ASR9K, and they always
> recommended using different silicon if you want to implement BGP ORR.
>
> At that time, we couldn't run BGP ORR because we had only 9k's, which was
> out of the question due to their ongoing issues. Now that it's time to
> replace the 9k with something else, I wanted to get your thoughts.
>
> Cheers,
>
> Darko
> _______________________________________________
> 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/



-- 
  ++ytti