|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
From: Subhasree Mandal (sumandal
NORTELNETWORKS.COM)Date: Wed Apr 25 2001 - 14:58:32 CDT
Interesting scenario, if Subhasree actually wants R1 and R2 to *not* form an
adjacency.
This is exactly the scenaio I wanted to know about.
In this case however, the two network LSA's for N1 that R0 recieves, will be
completely differenent.
R1's network LSA will contain information about routers on N1 that have
interfaces configured in area 1.1.1.1 and R2's will have info about 2.2.2.2.
I agree upto this point. Now let us say we do not have any other ABR in
area 1.1.1.1 or 2.2.2.2. R0 is the only ABR (and it is directly connected
to back-bone also)
Now when R0 computes intra-area SPF for area 1.1.1.1, it would not know
about the lsa in 2.2.2.2 and vice-versa. If the existing route belongs to
some other area, should it consider as an area change for the route and keep
toggling between two areas as we compute intra-area SPF for the two?
The options I can think of are:
1. Compare the cost against the existing route, and replace only if lower
But the problem is, it may happen that the existing route will not be there
any more when we complete SPF for the other area. In that case I will end
up with a woung routing information at the end of the complete SPF for all
areas.
2. Replace it if the existing one is not from the same SPF calculation,
i.e., it is from last time I calculated SPF for the other area I hadthis
route. This would give me the lowest cost route and the end always, but the
problem here is, I may toggle the route from area 1.1.1.1 to 2.2.2.2 and
back to 1.1.1.1 every time I run SPF. (or the other way, depending on the
seq of areas and cost) == unstable routing table.
3. Replace it every time I see the same route in the routing table from
another area, i.e. consider it as a area change for the route. In this case
again I may end up having a higher cost route even though there is a
possible lower cost route through the other area.
So is there a better way to deal with this situation, if we allow this kind
of configuration?
R0 will not have multiple routes to routers on N1(not via the network lsa
at least), unless N1 has another ABR with at least two interfaces on N1, one
in 1.1.1.1 and the other in 2.2.2.2.
The route to such a router will be selected based on cost.
In all of this I am asuming that the ABR's are configured properly to either
have an interface on the backbone, or a virtual link to it.
Nikhil Sama,
Nexsi Systems Corp.
408 579 5880
Philip Chen wrote:
Nikhil Sama wrote:
The network N1 can be part of only one area.
If R1 and R2 are independently configured to have their respective
interfaces on N1 to be part of area 1.1.1.1 and 2.2.2.2 respectively, they
will not form an adjacency, since the area field in the Hello packets will
not match.
Correct!
I do not think this was the intention, hence this would not be a valid
configuration.
You would have to ask Subhasree if this is his intention or not. If this is
an valid configuration or not largely depends on the answer to if this is
his intention or not and what he want to achieve via this configuration.
If however N1 is part of one area, then either R1 or R2 must be an ABR.
Right!
In this case R0 would reciecve two routes to N1, one intra-area route and
one inter area route. Since intra-area routes are always preferred, that is
the route that will be used.
R0 will not see N1 as an inter-area route unless one of the two conditions
(I mentioned earlier) is met by R1 or R2 whichever attached to two area.
--Phil
Nikhil Sama
Nexsi Systems Corp.
408 579 5880
Philip Chen wrote:
Subhasree Mandal wrote:
According to you it is a valid configuration. Which path should the ABR
choose?
The one with smaller path cost. If the cost are the same and R0 support
ECMP. R0 can have two paths to N1.
Note: Neither R1 nor R2 can reach N1 via R0 in the case of link failure
between R1 or R2 and N, unless you have one of these:
1) R0 has connectivity to backbone area 0.0.0.0
2) R0 supports alternative ABR behaviour
--Phil
Subhasree
-----Original Message-----
From: Manral, Vishwas [ mailto:VishwasM
netplane.com
<mailto:VishwasM
netplane.com> ]
Sent: Tuesday, April 24, 2001 9:00 PM
To: Mandal, Subhasree [SC7:370:EXCH]
Subject: RE: A Configuration question
ya, it is very valid !!!! but yes routers are where boundaries of areas are.
(The ascii diagram hasnt comeout too fine)
-----Original Message-----
From: Subhasree Mandal [ mailto:sumandal
NORTELNETWORKS.COM
<mailto:sumandal
NORTELNETWORKS.COM> ]
Sent: Wednesday, April 25, 2001 8:46 AM
To: OSPF
DISCUSS.MICROSOFT.COM
Subject: A Configuration question
Is this a valid configuration
------------------------------N1
area | |
1.1.1.1 | |area 2.2.2.2
R1 R2
| |
---------R0-------- area 2.2.2.2
area 1.1.1.1
So that router R0 will learn about network N1 from both areas?
--Nikhil Sama E-Mail: Nikhil.Sama
nexsi.com
Routing Protocols Group Ph: 408 579 5880
Nexsi Corp. Fax: 408 579 5701
1959 Concourse Drive Ph: (M) 310 428 9191
San Jose, CA 95131. www.nexsi.com
Home:
1240 Dale Av., Apt 16 Ph: 650 237 9055
Mountain View, CA 94040
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]