04-20-2006 08:10 AM
1)why is it mandatory that in targeted ldp we need to use loopbacks only for peering..?why cannot I use one of the physical interfaces.
2) In which all scenarios/services we use targeted LDP.AS far as I know only VPWS services..Whether targeted LDP can be used in MPLS L3 vpns..?
Thanks in advance
Thomas.
04-20-2006 08:41 AM
1) This is the way the system has been implemented. In which way does it restricts you?
2) If you want LDP over MPLS TE Tunnels then a directed LDP session is also used between the head end and the tail end.
Hope this helps,
04-21-2006 05:35 AM
Harold,
Thanks for your reply..I was just thinking abt the possibility of LDP over LDP(dont know whether it is possible).something like 2 P.E nodes connected back to back with no P nodes in between,and having a targeted LDP session and direct LDP neighbor relationship at the same time to the same peer P.E.
Just wondering whether these type of implementation is supported.?
2) Also as I asked in my earlier psot,whether targeted LDP can be ever used in MPLS L3 VPNs..?
Thanks in advance
Thomas.
04-21-2006 06:06 AM
Hello,
yes it is supported to have targeted LDP and "mpls ip" on an interface connecting two MPLS routers. There will be only one TCP session resulting from this scenario.
Targeted LDP can be configured manually and is useful in a MPLS L3VPN environment to avoid link recovery issues. In case a link fails and is restored, there is a race condition between the IGP and LDP. in the not unlikely case that the IGP converges faster than LDP establishes a TCP session and advertises labels, a path could be selected by the IGP, which does not have an LSP yet. Thus there is an outage for MPLS L3VPN traffic.
This can be avoided through targeted LDP, because the session is still active even if the direct link between two MPLS routers fails. Thus after link recovery the LIB still contains the LDP labels of the neighbor and thus an LSP is established as soon as the IGP converges.
Hope this helps! Please rate all posts.
Regards, Martin
04-21-2006 06:21 AM
Nice explanation Martin.
The targeted session is configured using the command
mpls ldp neigh
04-21-2006 07:31 AM
martin,
Thanks for your explanation..But I am afraid I have to disagree with your last sentence given below.
<<
My thinking is that the scenario is LDP over LDP,because in the targeted LDP the source and the destination will be the loopbacks and the other it is physical interfaces.So technically it is LDP over LDP..
Any comments..? or correct me if I got it wrong..
04-21-2006 07:41 AM
Hello,
I guess you are wrong as long as we consider the default behavoiur of LDP in frame-mode MPLS. By default LDP (or TDP) hellos are sent on every interface configured with "mpls ip". These hellos contain the LDP ID, which is in the form of
And this IP is used for the LDP TCP session. So it will not be "LDP over LDP". There will be only one TCP session to port 646 between the loopback addresses of neighboring MPLS routers.
You could reconfigure a MPLS router to use the physical interface address for LDP, but this is usually not done, because you might loose all LDP sessions because of an interface failure.
There is also no gain in trying to force more than one LDP session, because there is only one platform-wide label space in frame-based MPLS. Once a label is advertised it doesn´t make sense to advertise it again.
Hope this helps! Please rate all posts.
Regards, Martin
04-21-2006 07:55 AM
martin,
Thanks for your explanation..But I am afraid I have to disagree with your last sentence given below.
<<
My thinking is that the scenario is LDP over LDP,because in the targeted LDP the source and the destination will be the loopbacks and the other it is physical interfaces.So technically it is LDP over LDP..
Any comments..? or correct me if I got it wrong..
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide