----- Original Message -----
Sent: Saturday, November 08, 2003 8:08
AM
Subject: Re: [MPLS-OPS]:
CSPF-C=SPF?
At 10:00 PM 11/7/2003 -0800, Carlos Patriawan wrote:
----- Original Message -----
From: "Jean
Philippe Vasseur" <jvasseur@cisco.com>
To: <Devendra.Vyas@relianceinfo.com>
Cc:
<mpls-ops@mplsrc.com>
Sent:
Friday, November 07, 2003 5:45 AM
Subject: Re: [MPLS-OPS]:
CSPF-C=SPF?
> At 06:32 PM 11/7/2003 +0530, Devendra.Vyas@relianceinfo.com
wrote:
>
> >in the abscence of
> >*Administrative
groups (that is, link color requirements)
> >*Priority (setup and
hold)
> >*Explicit route (strict or loose)
> >i.e.,
without any constraints>>
> >is the CSPF path=IGP
path??
>
> if you do not have any constraint, yes, CSPF will
select the IGP shortest
> path, with the appropriate metric (could be
the TE metric or the IGP
metric)
>
> JP.
>
This
is implementation specific.
Some vendor may force unconstrained cspf
output to be equal with IGP best
path,while other vendor can give
different path for load-balancing reason,
for example if least-fill or
most-fill link bw is used in cspf computation.
The question, "if you do not have any constraint" then in this case,
=============
Yes,but that's LSP
constrain.
Even if lsp has no constrain,IF one vendor take TE-LINK BW
as a factor to decide cspf
output for lsp load-balancing reason,the cspf
result MAY NOT be always the same with
IGP best
path.
For example,if cspf found 3
best path,it can choose any path that has the least BW
used.Lets assume the best IGP path link only
has 20Mb left while alternative link1
(which is not best
IGP path) has 100Mb and alternative link2(again not the best
path)
has 200Mb left,cspf may choose link 2 as final result if "least-fill" is
implemented.
The reason why others implement it like this is simple,if
there're 1000 or multiple lsps
from headend to tailend and there are many
alternative path,why pushing all LSP
to take single link? It consume too
much resources and one single node failure cause
all lsp to go
down.
Thanks,
Carlos