[c-nsp] Difference in IP FRR Link vs Per Prefix Option

[Adam Vitkovsky]

Hi Amit,
Well I believe you can look at it as link-protection vs node-protection. 

Per-link LFA provides only link-protection -that means protecting node (node
computing the backup path for its primary links) will only chose LFA
node/candidate that has a link to primary next-hop node (node currently
acting as primary next-hop). 
In other words it will compute backup path that has to end on the same node
currently acting as primary next hop. 
And as you can see this might not always be the case in your topology as the
LFA candidate node might not have a direct link to your primary next-hop
though it might have a link to some other upstream node (node on a loop free
path towards the destination). 
Also 
Per-link is essentially per-next hop LFA (by next hop I mean the other end
of the directly connected link) so the protecting node computes only one
alternate/repair next hop for each primary next hop. The consequence of this
is that all prefixes with the common primary next-hop share also the same
backup next-hop. And there can be cases where the protecting node would have
multiple lower BW links to one/more LFA candidate nodes, instead of one link
with BW equal to the primary link. In such a case per-link LFA would chose
only one of the links and in case of primary-link failure the one lower BW
link might get congested. The congestion might as well occur on primary and
backup link with equal BWs as the link serving as backup might already be
carrying some level of traffic on its own and might not cope well with the
additional load during the primary link failure. So I case the protecting
node found more LFA candidates for a given primary next-hop it would
consider just one of them. 

Per-prefix LFA provides node-protection that is, the protecting node will
consider also LFA candidates that do not have a direct link to node acting
as primary next-hop however they have a link to an upstream node. This
increases the possibility of finding the LFA candidate(s) as well as
provides protection in case the node currently acting as primary next-hop
fails completely. 
And
In addition to node-protection, repair node running Per-prefix LFA computes
LFA next-hop per each prefix independently. That means if the protecting
node finds multiple nodes qualifying as LFA candidates for a given primary
next-hop, in case of primary next-hop failure it will load-share the traffic
among multiple backup paths. 


It's still morning for me so I hope it all makes some sense. 

adam
-----Original Message-----
From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of
Dhamija Amit
Sent: Tuesday, June 04, 2013 8:44 AM
To: cisco-nsp at puck.nether.net
Subject: [c-nsp] Difference in IP FRR Link vs Per Prefix Option


 Hi 

Can some one help me in understanding the exact difference in IP FRR Per
Link and Per Prefix option , It's mention in some docs that coverage of
protected prefixes will be more in per prefix option , If we have per link
then all the prefixes learned through same link will be covered in LFA
computation then what else is left for any router the IGP prefixes will be
learned through some link and if link is protected means all prefixes are
protected then how can we say coverage is more in per prefix.

I see some vendors Juniper , ALU doesn't supports per prefix option.Could
you please elaborate more on this.

Thanks for help

Regards
Amit Dhamija

_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/