[pim] Re: PIM vs. mLDP MT/FA

[[e-mail]]

Hi Peter, 
Thank you!
We'd like to find the most suitable way to use FA in PIM. 
I will read RFC9502 to see if it can be used.
Best regards,
Sandy











Original


From: PeterPsenak <[e-mail]>
To: 张征00007940;
Cc: [e-mail] <[e-mail]>;[e-mail] <[e-mail]>;
Date: 2024年07月26日 16:29
Subject: Re: [pim] Re: PIM vs. mLDP MT/FA



Hi Sandy,
 
On 25/07/2024 23:31, [e-mail] wrote:
 


Hi Peter, 
Thank you for your explanation!
The PIM is not binded to SR. 
 


Flex-algo can compute path to:1. SR-MPLS prefixes with algo specific label - SR-MPLS data-plane
2. SRv6 locators - SRv6 data-plane
3. IP algo prefixes - RFC 9502. - IP algo data-plane
In addition FA needs routers to announce participation in one of the specific data-planes from the above list.



PIM needs the unicast next hop calculated by IGP in specific MT or FA to the Source, but the label or SR locator isn't needed. 
 


I'm afraid FA can not give you what you want at the moment. Maybe we can extend it to do so ...
 



So whatever the Group is, PIM just lookup the IGP table to get the next hop to the source. 
 


it's not that simple with flex-algo :)
thanks,
 Petetr
 



PIM still can get what needed even if the Source is the same but the Groups are different.
Thanks,
Sandy
 
 



 




From: PeterPsenak <ppsenak=[e-mail]>
To: Jeffrey (Zhaohui) Zhang <[e-mail]>;
Cc: '[e-mail]' <[e-mail]>;
Date: 2024年07月25日 05:55
Subject: [pim] Re: PIM vs. mLDP MT/FA

_______________________________________________
 pim mailing list -- [e-mail]
 To unsubscribe send an email to [e-mail]
 
Hi Jeffrey ,
 
On 24/07/2024 23:42, Jeffrey (Zhaohui) Zhang wrote:
 
Hi Peter, Thanks for the reminder that in the case of mLDP and SRv6-P2MP the labels/locators can be tied to MT/FA. In the case of PIM, we need to clarify that an (s,g) tree will be tied to a particular MT/FA - not that there could be multiple per-MT/FA trees for the same (s,g). the point is that in  SRv6 there is only a single path to any locator, which is in an algo to which the locator belongs. There could not be any other algo path computed for it. 
Coming back to your case, the path would then be per source (assuming it is covered by the locator) and can not be per (s,g). For different algo you need a different source in SRv6.
If you want to be able to have paths in different algo for the same prefix, you can use SR-MPLS, or we would have to come up with something new....
thanks,
 Peter
 

Thanks. Jeffrey Juniper Business Use Only