08-13-2009 10:05 PM
In some scenario, I can see the use the RSVP for the mpls LSP while in other scenario, LDP is used for mpls LSP.
Really want to know which one is preferred in the real environment and what is the different.
08-14-2009 12:43 AM
here are the major differences (if we are talking about transport tunnels):
1) RSVP is the only choice if you need Traffic Engineering (FRR, link coloring, DiffServ-TE, Explicit Routing etc).
2) LDP is easier to configure (less provisioning). You just need to enable it on the interfaces. Label bindings automatically established for specific FECs (behaviour is different by different vendors).
08-14-2009 12:45 AM
From what I see, it depends a lot on your network.
If you want to benefit for fast reroute etc then you must go with RSVP. It also gives you the opportunity for next-gen multicast with "provider tunnels"
However the problem with RSVP is it requires fully meshed LSP's between all PE's hence you get the n-squared issue.
Some folks don't like this as it creates 100's if not 1000's of LSP's.
A nice alternative is large networks, is the P-PE connectivity is done with LDP, and LDP is tunneled through RSVP between P-P .
Therefore you have a very easy setup at PE-P level, whilst enabling Fast reroute between all P routers.
I would say, in a "small" network you may as well go with RSVP ... however if you've 100's of PE's, you may need to reconsider.
08-14-2009 12:47 AM
I guess the classical answer "it depends" is the only possible here.
LDP is probably the way to go if you want to have a fast start on MPLS with minimal configuration and decisions to make. The protocol, when enabled on all core interfaces, will automatically build the label switched paths from all possible ingress points to all possible egress points.
RSVP on the other hand offers more control, but this comes at the expense of more configuration statements being required. Indeed, you need to manually configure all required label switched paths on each of the ingress nodes. Hence the overhead depends on the number of PEs in your network. The main reasons why you would opt for it are traffic engineering (you decide where the traffic goes, with much more granularity than playing with the IGP metrics) and fast restoration (the typical goal of 50ms recovery when something breaks in the core).
I hope this helps.
08-14-2009 12:56 AM
As I do most LSP by RSVP, I don't realize the LDP can support LSP (manually specific LSP, not IGP) as well.
Seems that LDP can support manually LSP also.
So may I make one conclusion: RSVP is more useful only for FRR?
08-14-2009 01:07 AM
the LDP referenced in that page is targetted LDP, so it's not for transport.
If you look at the configs, you see they are using RSVP for transport.
Today, you can only do TE with RSVP. LDP will always follow your IGP.
08-14-2009 02:01 AM
you're right. The specific "protocols rsvp" configuration is missing. The "protocols mpls label-switched-path x" will only work with rsvp.
I'll have it changed in the manual.
07-22-2011 01:21 AM
Apart from the reserved MPLS labels; what range of MPLS labels are reserved specific protocols
label range for LDP -- label range for RSVP --label range for BGP --
By looking at the label, is there a way to find out protocol used such as LDP/RSVP/BGP MP
07-22-2011 02:32 AM
> By looking at the label, is there a way to find out protocol used such as LDP/RSVP/BGP MP
But, of your problem is just to find out how you learnt a given label..
what about "show route table mpls"?