Jun
28

In the previous MPLS Components post, we discussed the many benefits that MPLS can bring to the network, and we detailed the typical components found in a Layer 3 MPLS VPN design. In this post, we will provide more details for the MPLS components and their important, inner workings. We will make reference to the previous diagram in this post as well:

MPLS Components

MPLS Components

When PE1 receives a packet from CE1, it will engage in what we call a Push operation. PE1 is considered the ingress PE router and engages in label imposition. (Notice that we like to speak in fancy terminology here; when we add a label to a packet, it is termed a push or an imposition).

The P routers in the scenario will move the packets by simply swapping labels. How are the labels used in the Label Switch Path (LSP) learned by all of the routers? This is the job of the Label Distribution Protocol, or other existing protocols, but that is for later blog posts.

At the egress PE2 device, we have label disposition, or what we call a Pop of the label. (Fancy language for the removal of the label). If the second to last device in the path removes the label for us, this is termed Penultimate Hop Popping (PHP) and is the default Cisco implementation.

So we have pointed out that our example relies upon the Label Distribution Protocol (LDP) for the assignment of labels through the Label Switch Path (LSP). But how does LDP assign these labels? On what does it base its information?

It turns out that LDP relies upon the underlying IGP to build the best path for the LSP through the network. In fact, it also relies upon the IGP for loop free pathing.

This relationship between LDP and the IGP has many interesting aspects. For example, if the IGP reconverges on a new best path, so will the LSP through LDP. If there is a loop created or a blackhole situation created by the IGP reconvergence, this will also impact the LSP. Also, consider convergence times. LDP is certainly bound by the convergence time of the underlying IGP. Finally, consider the fact that this reliance brings up the need for inter-AS MPLS mechanisms for LDP.

The last point I want to discuss in this part is the fact that we often have label stacking with MPLS. In the case of our Layer 3 MPLS VPNs in the R&S track, the outer label (or transport label), is used to move the packet through the LSP, while the inner label is used to identify the VPN site. This is often called the VPN label.

Well, I certainly hope you are enjoying the posts on MPLS and there will be many more to come.


You can leave a response, or trackback from your own site.

10 Responses to “MPLS Components, Part 2”

 
  1. MO says:

    Hi,

    I didn’t understand the last paragraph:

    “The last point I want to discuss in this part is the fact that we often have label stacking with MPLS. In the case of our Layer 3 MPLS VPNs in the R&S track, the outer label (or transport label), is used to move the packet through the LSP, while the inner label is used to identify the VPN site. This is often called the VPN label.”

    Can you explain this more please?

    Thanks

    • @MO

      Sure! The PE router assigns a label for the VPN prefix. This is the inner label. It also assigns a label for the next hop PE router. This is the outer label (or transport label). The P routers in the middle of the cloud just act upon the outer label. They swap this outer label as they move the packet to the next PE hop. With PHP (Penultimate Hop Popping), the second to last device in the path pops off the outer label and hands it to the PE, which conveniently now has the inner label exposed for the proper VPN handoff.

      I will be sure to do a blog that demonstrates all of this.

      Thanks so much for reading and participating in the INE Blog!

  2. MO says:

    Thanks.

    So will the Ingress PE use the same outer label for a specific next hop PE router regardless of the inner label / prefix?

    - MO

  3. Raheel says:

    I’d love to see a blog on L2VPN as well.

  4. @MO
    The ingress PE determines the next hop for the prefix, then creates an inner label for the VPN prefix. It then creates an outer label for the appropriate next hop to move the packet through the core using the outer label only.

  5. Js says:

    It sound for me, works as below:

    Inner label = PE1 to PE2
    Outer label = P1 -> P2 -> P3

    Conclusion: Label stack

  6. Prathap says:

    PE-P-P-PE SP setup with multi vendor running IGP as OSPF have issues related to LSP’s flapping.

    These LSPs are flapping randomly as well as whenever there is a change in the topology.

    All the protocol measures and convergence mechanisms like efm, bfd are all deployed in place. LSP are RSVP_TE based LSP.

    I have narrowed to something which i can see as an issues i.e SPF timers which are varying in PE & P.

    Can anyone tell me whether this causes the LSP to flap either randomly / whenever topology changes.

    I highly appreciate the answers and please share some document to understand about this.

    Thank you very much,

  7. Roy says:

    I wanted to know, which router will impose the TRANSPORT label and which one will impose the VPN label ? and also, which routers shall dispose them at the other end ?

 

Leave a Reply

Categories

CCIE Bloggers