Segment Routing over IPv6 (SRv6)
Explain the Segment Routing architecture, the SRH, and SRv6 network programming with SIDs and behaviours. Configure SRv6 with an IGP (IS-IS or OSPFv3 extensions). Build an SRv6 L3VPN overlay service.
Take the quiz · pass mark 70% · 9 questions
Explanation
Segment Routing (SR) encodes the path a packet should take as a list of instructions in the packet itself, so the network core needs no per-flow state. SRv6 is SR using IPv6 addresses as the instructions — which is why it belongs in an IPv6 course.
The architecture (RFC 8402)
RFC 8402 defines the SR architecture. A segment is an instruction, identified by a Segment Identifier (SID). In SRv6 a SID is a full 128-bit IPv6 address, so SIDs are routed by the ordinary IPv6 control plane. The source (or an ingress router) imposes an ordered segment list; each segment is executed in turn. There are two SR planes; SRv6 is the IPv6 data plane, where the segment list travels in a new routing extension header.
The SRH (RFC 8754)
The Segment Routing Header (SRH, RFC 8754) is an IPv6 Routing extension header (type 4) carrying the segment list (an array of 128-bit SIDs), a Segments Left counter, and optional TLVs. As the packet visits each segment endpoint, Segments Left is decremented and the IPv6 destination is updated to the next SID. Because each SID is 16 bytes, a long path is a real header cost (see the worked example), so keep segment lists short or use compression (uSID).
Network programming (RFC 8986)
RFC 8986 defines SRv6 network programming: a SID is not just an address but an address bound to a behaviour (a function). The SID is structured as Locator : Function : (Argument) — the Locator routes the packet to the node, and the Function tells that node what to do. Standard behaviours include End (plain transit to the next SID), End.X (forward out a specific link), and the VPN behaviours End.DT4 / End.DT6 / End.DX4 / End.DX6 (decapsulate and look up / cross-connect into a VPN table). This "network as a computer" model is what makes SRv6 powerful.
Control plane and services (RFC 9352, 9513, 9252)
SRv6 SIDs and locators are advertised by the IGP: IS-IS extensions (RFC 9352) or OSPFv3 extensions (RFC 9513) distribute the locators so SIDs are reachable. Overlay services (L3VPN/EVPN) are signalled with BGP (RFC 9252), which advertises the service SID (e.g. an End.DT4/DT6) so a remote PE can send VPN traffic straight to it. Together: the IGP makes locators reachable, BGP binds a service to a SID, and the SRH steers packets through the chosen segments.
ictNetworks scenario
ictNetworks runs SRv6 in its core: locators from its documentation space are advertised by IS-IS (RFC 9352), traffic-engineered paths are expressed as short segment lists in the SRH (RFC 8754), and customer L3VPNs use End.DT6/DT4 service SIDs signalled by BGP (RFC 9252). It keeps segment lists short (and evaluates uSID) to control SRH overhead, and it filters SIDs at the domain edge so the SRv6 domain is not externally addressable.
Drawing diagram…
Worked examples
Example 1: A SID is an address
An SRv6 SID is a full IPv6 address, so it is 16 bytes (128 bits). That is why SIDs are routed by the normal IPv6 control plane and why a segment list is not free.
Example 2: SRH overhead for three segments
An SRH has an 8-byte fixed part plus one 16-byte entry per SID. A segment list of 3 SIDs therefore adds 8 + 16 * 3 = 56 bytes of header on top of the IPv6 header — a real MTU cost that motivates short lists and uSID.
Example 3: Splitting a SID
In SRv6 network programming a SID is Locator : Function : Argument. With a /64 locator, 128 - 64 = 64 bits remain for the Function and Argument, which is how one node can expose many distinct behaviours (End, End.X, End.DT6, …) under one locator.
Lab
Goal: Steer traffic over an explicit SRv6 path, then add an L3VPN.
You need: a containerlab topology with Linux seg6 (kernel SRv6) and FRR on four routers, plus hosts. (SRv6 support and syntax vary by kernel and FRR version — check what your versions implement.)
- Enable SRv6 on the routers, assign locators, and advertise them with the IGP (IS-IS or OSPFv3); confirm SIDs are reachable.
- From the ingress, impose an explicit segment list (SRH) steering a flow along a non-shortest path; capture the SRH and watch Segments Left decrement.
- Measure the MTU impact of the SRH and confirm PMTUD/fragmentation behaviour.
- Add an L3VPN: assign an End.DT6 (and/or End.DT4) service SID on the egress PE and signal it with BGP (RFC 9252); send VPN traffic and confirm it is decapsulated into the right table.
- Filter SIDs at the domain edge and confirm the SRv6 domain is not reachable from outside.
What to submit: the IGP/BGP/SRv6 configuration, the SRH capture showing the explicit path, the MTU note, and the working L3VPN with evidence of the service SID.
Summary
- SRv6 encodes the path as a list of IPv6-address SIDs in the SRH (RFC 8402, 8754); the core keeps no per-flow state.
- SRv6 network programming binds behaviours to SIDs (Locator:Function:Argument): End, End.X, End.DT4/DT6, etc. (RFC 8986).
- IGP extensions advertise locators (IS-IS RFC 9352, OSPFv3 RFC 9513); BGP binds VPN services to SIDs (RFC 9252).
- A SID is 16 bytes, so an SRH with N SIDs costs 8 + 16N bytes; keep lists short and filter SIDs at the edge.
- Next lesson (L8.02): IPv6 multicast routing.
Quiz imported · 9 questions · pass mark 70%. Lab questions are not auto-marked.