MPLS

That statement doesn't make sense without a lot more explaination behind it because without a context, it could mean just about anything.

I read it as control plane.

Sadly I can't think how to explain right now. And I may be wrong, again.

Hi Thrill5 and bod43,

Let's consider an IP/MPLS network. If MPLS uses IP routing protocols, we can't say that MPLS is connectionless, like it is for IP.

(1) What's the main difference between OSPF and OSPF-TE ?

(2) MPLS can setup its own path setup once it has the IP routing map. Do you know why?

Best regards, Michelot

Hi Michelot,

It's probably for a PE which has an LSR function together with an IP router function. For a pure LSR which doesn't terminates the IP path, the LSP path set up doesn't need to run firstly an IP routing protocol.

It seems there are ambiguous statements about that.

Best regards, Michelot

In a normal MPLS network today the LSP path setup is done either by LDP or by RSVP. These both run on top of IP.

It would be perfectly possible to setup LSP paths in a different way. For instance: Some types of MPLS equipment let you configure static MPLS paths - which are then somewhat similar to static IP routes.

But the standard way of doing it is via signaling protocols (LDP and RSVP) that run on top of IP.

Steinar Haug, Nethelp consulting, snipped-for-privacy@nethelp.no

Hi, Steinar

Happy to read you

Thanks for reminding me that. Yes, we don't have to forget that LDP or RSVP run on the top of IP, through UDP or TCP.

So, before to setup an LSP path, each LSR equipment has to build its IP routing table, using an IP routing protocol, probably BGP or OSPF.

Thanks for this remark

With your comments, we certainly have to be more precise to distinguish an LSR router when it plays a role of PE or P equipment.

- As PE equipment, the LSR considers all the IP datagrams, it terminates the IP traffic paths and the IP signaling paths,

- As P equipment, the LSR doesn=92t consider all the IP datagrams. The IP traffic paths are not terminated, but the IP signaling paths are terminated when the destination IP address is those of its own signaling interface.

Now, I=92m wondering how we can distinguish MPLS packets carrying IP traffic packets from MPLS packets carrying IP signaling packets. Perhaps with the number of the label (perhaps #14).

Best regards, Michelot

Packets for the various IP signaling protocols (IS-IS, OSPF, BGP) and MPLS signaling protocols (LDP, RSVP) are normally sent unlabelled - i.e. they are *not* sent as MPLS packets. IP packets and MPLS packets can normally be distinguished by at the link layer by checking the link layer type (i.e. for Ethernet: IP = 0x0800, MPLS = 0x8847). And IS-IS, of course, does not run on top of IP at all.

Steinar Haug, Nethelp consulting, snipped-for-privacy@nethelp.no

Hi Steiner,

Thanks for these words.

Perhaps, we can have another behaviour, because MPLS packets are "stackable".

Through Ethernet e.g. the equipment discovers MPLS with the Ethertype. If S=0 in the MPLS header the packet is switched to the next hop. If S=1, the client packet is analyzed. For an IP datagram, the destination address is then analyzed.

If the IP destination address is those from the equipment, the payload is locally treated. It can carried LDP, RSVP, OSPF, BGP information.

If the IP destination address is not those from the equipment the IP datagram is routed to the next hop.

Finally, it is your description, but by adding MPLS levels. Do you agree?

Best regards, Michelot