Thanks to the SDN, once again, and the MPLS World Congress 2017 proves to be a significant networking event and a watershed moment in the controversy about segment routing between vendors and customers.
Segment Routing (SDN) has the support of the whole industry: We’ve been collaborating with lead operators and the Internet Engineering Task Force (IETF) since day one, and there’s a clear agreement among vendors that engage in Interop testings.
I’m quite impressed with how the Segment Routing discussion has progressed. Attendees are no longer debating whether or not they can use Segment Routing… They’re both sure it’s the right thing to do.
Now, the issue is all on the scope of the implementation. Segment Routing is no longer exclusive to Core networks; it has extended to Metro and Aggregation networks, as well as Data Centers and the Host!
You can set up end-to-end policies through your individual Metro, WAN, and Data Center realms with Segment Routing. This way, you will skip complicated protocol conversions between network realms while also preserving scalability – the epitome of simplification.
This is a game-changer since it was always impossible! What allows this to happen?
Next Hop On-Demand
Provisioning multi-domain services (L2VPN & L3VPN) is difficult and time-consuming, particularly when routing details must be redistributed throughout domains… There’s no need to redistribute something with the On-Demand Next Hop (ODN) function. ODN causes the XR Transport Controller (XTC) – a PCE controller – to delegate the computation of an end-to-end LSP, including restrictions and policies. It then stores the answered multi-domain LSP in the local FIB for the remainder of the operation.
You can effectively set up a path from the Top-of-Rack across your whole network infrastructure up to an Egress Peering router with restrictions like low latency or disjointness in this way.
On-the-Go Next Hop is an excellent illustration of why Segment Routing was created: to strike the right combination between dispersed knowledge and centralised optimization.
Support for Wide Segment Routing
We’ve had this end-to-end vision from the beginning, and we’ve decided to help Segment Routing through our whole range, which includes all IOS XR, IOS XE, and NX-OS hardware platforms.
And there’s more to come: Segment Routing IPv6 is gaining momentum in the market, and many IETF draughts have recently been released.
Segment Routing not only functions in a native IPv6 environment, but when used in tandem, it creates an entirely different paradigm.
Extension headers, such as Routing Extension Headers, have been built into the IPv6 packet header from the ground up. Without going into too much detail, think about Extension headers as a way to transfer instructions to each packet… You may simply guide your application on the right network route by using the Routing Extension headers to transfer Segment Routing IDs. In the IPv6 universe, however, the Segment Routing ID is 128 bits long! For destination addresses, you don’t need any of these bits… That leaves you with, say, 64 bits to point to a feature and move it some arguments!
Doesn’t that sound familiar? That should remind you of 101 programming courses, I’m sure.
Welcome to the world of Infrastructure as Code, as described by the modern paradigm.
If you want to learn more, watch the videos below, which were filmed at the MPLS World Congress 2017:
Microsoft’s Paul Mattes discusses segment routing for data centre interconnect at size.
Bell Canada’s Daniel Voyer discusses the restructuring of the central office.
Clarence Filsfils, Cisco, discusses deployment insight and technology updates.
Segment Routing is a novel architecture that was designed with SDN in mind. The goal of separating the data and control planes is to simplify the network. SDN is a fantastic idea, and it is now up to us to apply it to today’s networks. The SDN idea of simplicity is a driving force behind the implementation of Segment Routing.
Segment Routing employs the fundamentals of MPLS but with fewer protocols, less inter-protocol interaction, and less state, and it is applied to MPLS architecture with no changes to the forwarding layer.
Existing devices that use labels to switch may simply need a software update. The general idea is based on source routing. The route you travel via the network is determined by the source. It directs a packet via an organized sequence of instructions known as segments. Segment Routing, like MPLS, is based on label switching but lacks LDP and RSVP. Labels are now known as segments, and we still have push, swap, and pop operations. You do not maintain state in the middle of the network since the state is included in the packet.