1. You're corret.
2. You control that with your vrf import/export/both, and using route-maps
to filter the specific prefixes (regular BGP stuff)
3. The routes are shared with the BGP community values. The RD and
community values don't HAVE to be equal. All examples in the docs I gave
you the link to DO show them being equal. That's just a sanity check. You
can set the import community from any RD you feel like. But if you have a
big network, and make them all different you'll have some EVIL
administration to do all the time! (job security?)
4. Sanity
5. That's the "regular" LDP stuff. In the Cisco world, MPLS MUST HAVE CEF
running in order to work. So you'll have a fast "IP" lookup ability anyway.
But in an MPLS network, whenever a packet comes in, first a LIB (Label
Information Base) lookup is done, then a CEF lookup will be done if the LIB
has no entry. So in effect, no label, no route, therefore "regular" IP
routing concepts. If there isn't an entry, most likely the router will
pause, send an LDP request/query and establish a route and propogate a set
of labels through the network along a given path. That'll happen the first
time for every different type of flow (unless you traffic engineer and
pre-establish paths).
Scott
-----Original Message-----
From: Vinod Anthony Joseph Cherunni [mailto:vac@dsqworld.com]
Sent: Saturday, October 07, 2000 6:03 AM
To: eosborne@cisco.com
Cc: smorris@mentortech.com; cisco-nsp@puck.nether.net
Subject: Re: [nsp] REG: MPLS VPN route exchange
Hi All,
Thanks a ton once again for the great advice I have got, Since I am a
little new to the MPLS theory, I would like to get some more things clear
for our deployment, Pls hep me out.
1. My understanding is that members are grouped into a VPN, based on what
routes are available to them by means of the import / export attributes
"which point to a particular RD" & in effect populates the VRF which is
associated with a VPN. Kindly correct me.
2. Assume I have a customer site called A who needs to connect to his
branch office site called B & he wants both these sites to be part of the
same VPN called X. Another requirement of his is that site A needs to
connect to another partner called C (Extranet requirement) through a
seperate VPN called Y, wherein he wants site C to only be aware of a
particular IP Prefix of his site. I am assuming all these sites are
connected by means of two directly connected PE routers. How will I
configure the PE router to accomodate Site A to be part of two VPN's & also
ensure that site B gets all routes belonging to Site A, Partner Site C only
is only aware of a particular IP Prefix of Site A, & Site A has all the
routes of Site A & Site B
3. My next query is every Prefix is associated with an RD. Now say in my
network if the RD is kept to be only locally significant. There is always a
chance of a PE routers recieving routes with overlapping Address spaces
also associated with a same RD value. Kindly correct me.
4. What is the real advantage of having globally significant RD's?.
5. My understanding on the assigning of labels is as follows- Labels have
only local significance. Labels are only exchanged between directly
connected neighbors, Destinations are mapped to labels, & packets are
assigned with them at the ingress (PE). Now if a packet is tagged with a
label say "5" by a PE device, The meaning of this label would have been
exchanged with all the Core routers. Typically for routing this packet,
what kind of information or action needed to be taken would have been
exchanged between the PE & Core routers.
I am really thankful for all the help.
With kind regards,
Vinod
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:18 EDT