Dear Stewart,
thanks for comments, see minecomments inline.
On 07/02/2013 07:07 PM, Stewart Bryant wrote:
On 26/06/2013 14:13, Levente Csikor wrote:
Levente
In each case below the conditions are surely fulfilled if n=e
thus I think that in each case the condition needs to be
changed to :
some node n!={s or e}
LC> Yes, you have right that e should be excluded from the set of
examined nodes, but in the first case below, it does not surely fulfill
the condition,
LC> but it would conflict, since assume that n=e := X, then we would get
that
LC> dist(s,X) < dist(s,X) + dist(X,X), where obviously the last
component is zero!
LC> Moreover, if next-hop e (,which is not the destination) became
unreachable because of the failed link, then node s could not tunnel the
traffic to e
LC> again, since it would use the failed link. If we could do that, we
need one more label in the label stack, since we would need to reach
some other
LC> node t, then we would need a tunnel from node t to node e, and then
the remaining shortest path from node e to the destination.
LC> On the other hand, I think that node d has to remain in the set of
"non-analyzed" nodes, i.e., some node n!={s,e,d}, since according to the
draft, if
LC> we could tunnel the traffic directly to the destination (the case
when n=d) "then conventional LFA would have been sufficient to effect
the repair."
In order to not to read our full papers and searching the answers, I
copied here the relevant parts:
*Link-protecting rLFA condition:*
For source /s/, destination /d/, and next-hop /e/, some node /n/ /!=
s,d/ is a link-protecting remote LFA for the /s-d/ pair if and only if
/dist(s, n) < dist(s, e) + dist(e, n)/ (1)
/dist(n, d) < dist(n, s) + dist(s, d)/ (2)
In these equations, one can easily see, that (1) defines the P-space,
while (2) is the condition of Q-space. Furthermore, with these
formalized conditions, one can easily observe, that (2) is actually
the basic loop-free criterion of pure LFA.
*Link-protecting rLFA condition with extended P-space:*
For source /s/, destination /d/, and next-hop /e/, some node /n !=
s,d/ is an extended link-protecting remote LFA for the /s-d/ pair if
and only if
/?v ? neigh(s) : dist(v, n) < dist(v, s) + dist(s, e) + dist(e, n)///
/dist(n, d) < dist(n, s) + dist(s, d) . /
*Node-protecting rLFA condition:*
For source /s/, destination /d/, and next-hop /e/, some /n != s,d/ is
a node-protecting remote LFA for the /s-d/ pair if and only if
/dist(s,n) < dist(s,e) + dist(e,n)/// (3)
/dist(n,d) < dist(n,e) + dist(e,d) / (4)
As it was in the case of link protection, here, (3) defines the
P-space, while (4) characterize the Q-space.
Here, two important observations can be made, which are the followings:
- P-space does not depend on the protection scheme (i.e., link or
node protection)
SB> That falls directly out of the definition of P-space
SB> since in link you cannot traverse the link to e and in
SB> node only get to e if you traverse the link to e
SB> thus the exclusion of the link to e applies in both cases
LC> Yes it is kind of obvious observation, which falls out of the
definition of P-space, but I think that this could be noted in the
draft, since if someone tries
LC> to determine how much SPF,rSPF or even all-pair SP calculations are
needed, then it is good to remember that for P-space, we could deal with
the
LC> same amount of calculations in both protection cases.
- (4) again is the basic node-protecting loop-free criterion of pure
LFA.
*Node-protecting rLFA condition with extended P-space:*
For source /s/, destination /d/, and next-hop /e/, some node /n !=
s,d/ is an extended node-protecting remote LFA for the /s-d/ pair if
and only if
/?v ? neigh(s) : dist(v, n) < dist(v, e) + dist(e, n)/
/dist(n, d) < dist(n, e) + dist(e, d) ./
Despite the fact that we only considered unit cost networks, the
formal definitions above are *true for any arbitrary weighted network.*
SB> I do not see where the unit costs come into the text above. It looks
SB> like it is already expressed in terms of arbitrary cost.
LC> Since in our papers, we only dealt with unit cost networks,
therefore I wanted to emphasize that these results are true for any
arbitrary weighted network. (However, you have right about the fact that
if the conditions are expressed by distance functions, they will
definitely true with any link cost settings :) )
SB> Additionally in order to limit the number of SPFs to a practical
SB> level, we normally suggest that the repair target in not d, but
SB> instead is e (in the link case) or next hop of e (node case).
SB> Anyway I will work on some text.
LC> Thanks.
- Stewart
Regards,
Levente
_______________________________________________
rtgwg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/rtgwg
