[PDB-tech] proposed new attribute: max_allowed_peering_next_hop_latency

Kristian Larsson kristian at spritelink.net
Wed Apr 20 09:08:04 PDT 2016



On 19/04/16 17:32, Job Snijders wrote:
> On Tue, Apr 19, 2016 at 05:22:34PM +0200, Denis Fondras wrote:
>>> Thoughts?
>>
>> I have no usage for that field but if it is used by some peers, let's
>> add it.
>
> I can't speak for other networks, but for NTT Communications the peering
> next-hop latency is part of the public peering policy.

There are lots of requirements in the average peering policy, such as 
geographic coverage and whatnot but we don't encode that in predefined 
fields in pdb. What makes this value special?


>> Perhaps we could also add a "BFD" field stating if BFD is used by the
>> peer ?
>
> BFD is usually something negotiated/discussed on a per link basis, as
> such I don't think it should have a place in the network record. Also
> although the concept of BFD is simple, the actual implmentations and
> stablility of such implementations are far from mature.
>
> For me a key consideration in adding (or removing) fields in the
> PeeringDB data model is how they relate to network automation. I don't
> see myself being able to automate BFD sessions based on a simple
> (boolean) field.

You can enable BFD for all sessions and if the peer doesn't support it, 
it simply won't come up, right?! Is there any downside to this?

    kll


More information about the Pdb-tech mailing list