Re: [Idr] Clarifications on draft-ietf-idr-segment-routing-te-policy

["[e-mail]" <[e-mail]>]

Dear Ketan,

Understood. Is it proper to explicitly add some text in the doc like "RR Implementation is RECOMMENDED to provide a policy knob to only reflect related policies to relevant peers according to the route target attached to the policies" ?

I believe this feature is important for large scale deployment of SR policy.

Thank you very much for your consideration.


Best Regards,
Zhenqiang Li

[e-mail]
 
From: Ketan Talaulikar
Date: 2023-04-12 22:55
To: [e-mail]
CC: authors; [e-mail]; idr-ads; idr-chairs
Subject: Re: Re: Clarifications on draft-ietf-idr-segment-routing-te-policy
Hi Zhenqiang Li,

The deployment design that you describe can achieve the optimization that you desire by using outbound route policy on the RR to each of its neighbors. The spec does not prevent an implementation from providing a policy knob to achieve the behavior you desire.

Also, consider a deployment design with hierarchical RRs. Would your proposed change not break BGP propagation?

Personally, I do not think we should change the base BGP propagation mechanism for such scenarios. BGP is basically a p2mp distribution protocol and we need to be careful how far we bend it for other purposes.

Thanks,
Ketan


On Sat, Apr 8, 2023 at 6:51 AM [e-mail] <[e-mail]> wrote:
Hello Ketan,

I understand the objective and the current text do satisfy the objective. What I point out is the issue the objective has and the current text should be refined.

suppose a  RR has 100 peers, and 10 SR policies are to be configured on each peer. Follow the current text, RR will reflect 1000 SR policies to each peer and only 1% is needed. 



Best Regards,
Zhenqiang Li

[e-mail]
 
From: Ketan Talaulikar
Date: 2023-04-06 18:29
To: [e-mail]
CC: draft-ietf-idr-segment-routing-te-policy.authors; [e-mail]; idr-ads; idr-chairs
Subject: Re: Clarifications on draft-ietf-idr-segment-routing-te-policy
Hi Zhenqiang Li,

The objective was not to change the base BGP decision process or the BGP propagation rules for this SAFI. I hope that explains the current text.

Thanks,
Ketan


On Thu, Apr 6, 2023 at 1:53 PM [e-mail] <[e-mail]> wrote:
Hello authors and all,

Hope it is not too late to make clarifications on this draft. 

This draft says one or more IPv4 address format route target extended community attached to the SR Policy advertisement and that indicates the intended headend of such an SR Policy advertisement.
And a BGP node advertises a received SR Policy NLRI to its IBGP neighbors according to normal IBGP propagation rules. 

Here a BGP node can be a RR(Route Reflector) which will reflect all the SR Policies to its IBGP peers regardless of the route target attached to the policies. So the IBGP peer node as the headend will receive lots of policies that are not intended for it because those policies have no route target that indicates the received peer node. 

In my opinion, the rules for RR to reflect SR policy with route target SHOULD be enhanced. RR SHOULD reflect related policies to relevant peers according to the route target attached to the policies.



Best Regards,
Zhenqiang Li

[e-mail]