Hi Pierre,
agree, it does make sense to have local protection use case(s). We can get into
more details in requirements and draw conclusions with gap analysis doc.
Regards,
Greg
From: Pierre Francois [mailto:[email protected]]
Sent: Friday, April 04, 2014 2:17 AM
To: Robert Raszuk
Cc: Gregory Mirsky; [email protected];
[email protected]
Subject: Re: [spring] Mail regarding draft-francois-spring-resiliency-use-case
Robert,
Thanks for the clarification :)
Greg,
There are operators who want to have the network provide such protection.
You mind if I leave such approaches in the use-case doc, and we re-raise such
very relevant complexity/mess questions
when solution drafts are discussed?
Cheers,
Pierre.
On Apr 3, 2014, at 11:01 PM, Robert Raszuk
<[email protected]<mailto:[email protected]>> wrote:
Hi Greg & Pierre,
More seriously, yes, in SR, you have to pay attention to label allocation when
installing your failover entries in the FIB. However I do not see why it leads
to having intermediate nodes maintain path information.
Can you expand on this part of your comment?
GIM>> Link mode of local protection does not require any special consideration
as the MP is the same next-hop node of the protected link. But for node mode of
local protection special consideration must be given if SID functionality to be
used. Consider case when the next-hop node has advertised SID that is used by
some e2e paths that traverse the PLR and this node. As result, IMO, selection
of the MP depends not only on next-next-hop node of the SR path but on
availability of particular SID as well. Now we can make it more complex if the
protected node owns not one but several SIDs. I think that we may easily avoid
this complexity/mess by stating that only link mode local protection is
applicable to SPRING domains and leave service protection/redundancy to
Service/SFC OAM layer.
I think you are both right :)
I think what Greg you are refering to is state of the repair "paths" in regards
to IGP topology which clearly is required by all node protection solutions.
Contrary what I think Pierre consider as "path" is the end to end traffic flow
paths which would normally result in 100s or 1000s LSPs state - that clearly is
not required in SR node protection.
So what may be helpful is to rather then overloading term "path" here redefine
it for the purpose of this discussion into IGP topology state (as example).
Cheers,
R.
_______________________________________________
spring mailing list
[email protected]<mailto:[email protected]>
https://www.ietf.org/mailman/listinfo/spring
_______________________________________________
spring mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/spring
