0

If not utilizing any form of TE (Traffic Engineering), are there any benefits of using a transport label in L3VPN - instead of just using a simple IPinIP tunnel?

e.g. if traffic is egressing PE1 with a loopback address of 1.1.1.1 towards egress PE2 with a loopback address of 2.2.2.2, why not just encapsulate the packet inside another IP header with SIP:1.1.1.1 and DIP:2.2.2.2 instead of using labels?

1
  • 1
    That would be a different kind of VPN. Both are valid ways to create a VPN. – user253751 Apr 15 at 8:54
4

Two main reasons for using labels:

  1. It makes the transport layer independent of the customer addressing scheme. Multiple customers can have overlapping addressing. The transport doesn't care

  2. Protocol independence. What you've describe works fine for IP, but MPLS can handle non-IP protocols (L2, pseudowires) too.

3
  • Thanks! I agree with your second point. (But my question was specific to IP VPN, not L2 VPN). The part regarding overlapping addressing is taken care of by using internal VPN label not outer transport label, which only used for tunneling. I just don't know why is it better than other tunneling techniques in this specific scenario (or it has no benefits at all, just something commonly used?). When I first learned about MPLS VPN, I was told that the benefit is label lookup (ILM) is faster than IP lookup. But I think nowadays it's the same (since usually TCAM is used for IP lookup) – manish ma Apr 12 at 12:37
  • 1
    I guess the benefit is expendability. If I'll want to add L2 or IPv6 or some other services in the future it will be more simple. – manish ma Apr 12 at 12:50
  • Many protocol decisions were made at a time when limits on processing speed and memory were a significant issue. – Ron Trunk Apr 12 at 13:16
1

The SprintLink Peerless network is actually an IP-over-IP network, instead of a MPLS L3VPN network. Sprint engineers believed the IP routing technology to be mature and inexpensive, and viewed it as a better option than adopting new MPLS technology.

I mention this mainly to identify that there are different schools of thought on the available technology stacks.

In theory, MPLS should have brought us less-expensive backbone (P) nodes which didn't need a full IP routing table. This would've made networks look a little more like the IP-over-ATM and IP-over-Frame-Relay networks of the mid-90s and late-90s. However, in practice, vendors did not ship economical MPLS P boxes. Indeed, Cisco and Juniper both shipped numerous products that could be a PE but not a P; examples are many Cat 6500/7600 cards, and early EX-series switches.

Touching on some history, again in theory, an ATM or Frame-Relay switch or entire network could have functioned as an MPLS P node, being configured by a separate control-plane. To my knowledge, no such networks were ever deployed in production. However, the desire to offer this capability resulted in a number of MPLS protocols' design decisions, for example, per-interface label allocation.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.