MPLS technology and issues of adoption

Luca Martini has been primarily involved with L2VPN technology and pseudowires that evolved from his original draft-martini design. This technology, which is now an IETF standard track RFC, has been accepted as the de-facto industry standard transport for layer2 protocols over Multiprotocol Label Switching (MPLS).

He has been involved in the Internet engineering task force (IETF) for the past nine years, and contributed enhancements to the RFC2547 MPLS-VPN design. He is the author of the Draft-Martini design for transporting layer2 protocols over MPLS core network.

During an interaction with Prasoon Srivastava of CIOL, Luca touched upon evolution of MPLS technology and issues related to its adoption. Here are the excerpts:

How do you see MPLS technology evolving further?

I have created this technology called pseudowires many years ago with a team from Draft Martini. I really had no idea how far in time we would come and how popular would it become. Latest development is that people have realized old networks based on TDM/Sonnet were not a good fetch for IP technology.

What is happening now is that we have been leading in a new technology called MPLS Transport Profile, which is basically another addition of MPLS technology which we created 10 years ago.

Going forward, what I see is that, pseudowires, which are first client to the MPLS-TP technology, are going to take over the transmission work which I was never expecting. If you look at future, 10 years from now, we will probably see a world of transmission based on packets on MPLS with pseudowires as first client and IP as other client predominantly.

What are the key features that you feel will encourage its adoption over other technologies?

I think there are two main features. The first one is MPLS, a mutli-service platform which enables service providers to provide different services quickly. The second part is statistical multiplexing which is a feature of packet networks. In circuit-based network you have certain amount of dedicated bandwidth which is very wasteful.

Most of the circuits are empty. Packet-based network saves enormous amount of capital and expense. This is one great feature of packet network that will make it successful. If you have circuit-based network then it will be you only who will be using it but when it comes to packet-based network then you can share bandwidth with your neighbors.

Therefore when you are not using it somebody else can. This also allows service provider to over subscribe links. This is the biggest technology I believe is there in adapting packet technology.

How would you rate security parameters in MPLS TP than in existing MPLS?

Security will be no different than what we have in Sonnet. We separate Sonnet by channels, we separate MPLS by labels. We are talking about putting MPLS as a layer for transport network. One of the clients is IP and the other client is pseudowires; they still have separation. Level of security is pretty much the same as what we have today.

What is the status of ASIC (Application Specific Integrated Circuits) based routers in networking?

We have been running ASICs for years. In any network device, technologies have a certain progression. First implementations are made in software. Then you adapt with adding FPGAs (Filed Programmable Gate Arrays) and eventually you move to ASIC. This was the trend six or seven years ago, but now we are moving little quicker.

Today we probably skip software part because of the speed involved; we cannot any longer invest time in implementing software and then do FPGAs and ASICs. ASIC design and deployment cycle is fairly slow. It takes a couple of years. In Cisco, we have done a little better. We have designed some flexible ASICs that can easily adapt to new upcoming technologies. Therefore we don't have to design new ASIC every minute.

What are the issues that you identify, which come in way of deploying MPLS TP?

MPLS is a clever technology. It is neither layer 2 nor layer 3. It is layer 2.5. We have seen interesting trend in certain location with specific service providers. It is due to certain political organization in those countries where people say that it has to be layer 2 and it can not be layer 3.

For example, in India, interconnect between regions for voice has to be TDM layer2. This is legislative. Therefore, there are some service providers that might not move to MPLS TP right away because they may be stuck in something which has been labeled Layer 2.

You might have rings still based on Ethernet on SDH until some of those rules and regulations are changed to allow technology progress in the most efficient way. These kind of laws block at particular point of time.

Technology moves on but the countries are stuck because the law has not been changed there. They miss to keep pace with global technology.