Hi Jeff,
Thanks for confirming the flexible approach of Juniper on the topic of
multiple sessions and features.
To your point, my personal use-case, for example, would be to gain
visibility in valid non-primary paths. Primary paths i use to correlate
IPFIX against BMP data; valid non-primary paths for what-if analysis,
ie. capacity planning. Personally, i can stay rather flexible on the
topic of state compression in order to reduce churn as the work will be
more historical than real-time.
Paolo
On 3/4/25 01:58, Jeffrey Haas wrote:
Paolo,
On Apr 2, 2025, at 1:42 PM, Paolo Lucente <[email protected]> wrote:
On 25/3/25 15:47, Jeffrey Haas wrote:
[..]
Understanding the motivation potentially has some options. Perhaps this is
never turned on for applications using BMP for traffic shaping. Perhaps you
turn this on a "side session" that is only there for exposing this
information. Potentially such a side session uses "triggered BMP".
[ .. ]
The BMP & multiple sessions topic is something that returned also about the REL
draft; in that context, for example, the use-case was about two sessions, one to
get routes, one for debugging ie. log malformed BGP packets.
As an implementor of a BMP exporter, what would be your thoughts about making multiple
BMP sessions configurable at a BMP exporter, each with its own feature flags set on/off
(a generalization on your "side session" concept). So that, on top of defining
features optional in a draft, it can be further scoped where these features are needed in
practice.
Our implementation already permits subsets of data to be sent on a session. We
do have customers that utilize different sessions to convey different views
(slices) of the data.
We're already at the point where we've had discussions about applying policy to
those sessions where a given station may be interested in a smaller portion of
a given view.
Internally at Juniper and also among applications like IETF BGP YANG, we've had
discussions about what more targeted subscriptions might look like, and how
that might be signaled.
That said, it's better to not flail at the use cases for this. Our consequence of how
the marking works is that it is guaranteed to be chatty if it's done in a traditional BMP
streaming case. I'd encourage the authors and WG to consider what you actually want out
of the feature for a use case. I suspect a desire for "secret sauce" is keeping
people from being fully honest here. :-)
-- Jeff
_______________________________________________
GROW mailing list -- [email protected]
To unsubscribe send an email to [email protected]
