Hi Jeff,

Thanks for summarizing the groundwork and prior discussions.
I was planning to float one, with per-afi/safi and I started with the category 
which I shared and others sharing the same thought can contribute/collaborate.

I understand it can be noisy to club it with this one, but it could be helpful 
(and please provide your suggestion on this) in the longer run:

  1.  Call out in this draft about what we are not covering in this one.
  2.  Start individual drafts per AFI/SAFI (or just AFIs or subset of 
AFI/SAFI’s clubbed together) capturing its uniqueness. it will ensure that 
double clicking the “post-policy” can be delegated to each of these drafts.

Thanks,
Saumya.

From: Jeffrey Haas <[email protected]>
Date: Tuesday, 5 August 2025 at 8:57 PM
To: Dikshit, Saumya <[email protected]>
Cc: msri <[email protected]>, linchangwang <[email protected]>, 
[email protected] 
<[email protected]>, [email protected] <[email protected]>
Subject: Re: [GROW] Query on draft-ietf-grow-bmp-bgp-rib-stats
Saumya,


> On Aug 5, 2025, at 10:27 AM, Dikshit, Saumya 
> <[email protected]> wrote:
>
> I understand that pre-policy and post-policy is generic spectrum under which 
> all filtering is covered. Hence the query.
>
> Considering it’s BMP (aiding troubleshooting, monitoring the BGP Control 
> plane), there should be obvious and not-so-obvious sub-categories defined for 
> “post-policy”.
> Is this a work-done or parallelly being followed in the WG.

A general comment is that enhanced statistics is relatively new work in grow.  
If you were to review some prior working group meeting recordings, one of the 
things you'll see is some desire to not only add new vendor-neutral stats which 
this draft did, but find a release cadence for such new counters.  I.e., don't 
try to cram everything into a never-shipping internet-draft. :-)


> I was looking explicitly for cross EVI (EVPN) absorption facilitating Dynamic 
> IVRL in layer-3 and Cross-EVI MAC (collapsing bridge-domains) absorption in 
> case of layer-2.
> As I mentioned, was looking at VPN specific AFI’s including l2vpn and 
> vpnv4/v6.

I suspect you'll find members of the WG receptive to some additional per 
afi/safi stats.

For the EVI or more generally the per "instance" cases, this may get more 
interesting.  The general use case is roughly "how many routes have been 
learned by the instance?"  This further could break down to whether routes are 
learned within the instance, or have been leaked in via the VPN mechanisms.

The challenge is such statistics are not only per-router, but also for a high 
level definition of the "instance". Streaming such things abstractly in BMP 
will be an interesting keying challenge.  But it will also raise the question 
about whether the mechanism is "too noisy".  Discussion on the points should be 
interesting.



-- Jeff
_______________________________________________
GROW mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to