The big picture: why providers and data centres need new tools
What you will learn in this module. You will meet the whole Segment Routing and EVPN track before touching a single command. You will learn why a service provider wants labels without a separate label protocol, why a data centre wants an overlay with a real control plane, how the eight modules fit together, how the simulated labs behave, and a one-page map of every term you will meet later. Prerequisites: you know what an IP address, a subnet, a router and a switch are, and you have seen a routing protocol (OSPF or IS-IS) work. Nothing else is assumed.
An analogy: the railway and the courier
Imagine a country with a railway network. A normal IP network is like a train system where every station decides, for every parcel, which platform it goes to next, by reading the full address on the parcel. That works, but every station needs the whole map and every parcel is examined at every station.
Now give every parcel a short coloured tag. A station only reads the tag and puts the parcel on the matching train; it does not read the address. That is MPLS: a short label replaces the address lookup inside the provider network. The question is who tells the stations which tag means which destination. The old answer is a second protocol (LDP). The new answer, Segment Routing (SR), lets the routing protocol you already run (IS-IS) hand out the tags itself. One protocol fewer to run, one fewer thing to break.
Labels move the heavy work to the edge of the network.
Two worlds, one track
This track covers two connected worlds.
- The provider world (Segment Routing, L3VPN). A provider such as our simulated Kaveri Telecom carries many customers over one core. The core must be fast, simple and protected against link failures. SR gives the labels; an L3VPN gives each customer a private network on top.
- The data centre world (VXLAN, EVPN, ESI). A data centre has hundreds of racks. Virtual machines move and need the same subnet in different racks. VXLAN stretches a Layer 2 network over a Layer 3 fabric; EVPN is the control plane (based on BGP) that tells every switch where every MAC address and IP prefix lives; ESI lets a server connect to two switches at once without loops.
Both worlds share one idea: an underlay that is simple and an overlay that carries the service. The underlay only moves packets between loopbacks. The overlay says who may talk to whom. Once you see this pattern, SR with L3VPN and VXLAN with EVPN stop looking like two topics.
The 2 a.m. call
A provider engineer is told that a customer VPN is down. Pings between PE loopbacks work, so she assumes the core is healthy. In the lab you will see why that is a trap: plain IP works while labelled traffic fails, because the core has a router that forgot to switch on SR in IS-IS. She finds it by reading the label table, not by pinging.
Lesson: check the data plane the way the customer traffic uses it, with labels.
"Why would a provider replace LDP with segment routing?"
Fewer protocols: IS-IS distributes the labels, so there is no separate LDP session to maintain or to desynchronise. Labels are derived from an index and a common label block, so they are predictable. SR also enables TI-LFA fast reroute and traffic engineering without RSVP state in the core.
Key takeaways
- MPLS labels let core routers forward without customer routes.
- Segment Routing lets the IGP (IS-IS) distribute the labels, replacing LDP.
- VXLAN plus EVPN does for data centres what labels plus VPNs do for providers: an overlay with a control plane.
- Always think underlay (moves packets between loopbacks) and overlay (carries the service).