SRv6: Segment Routing over IPv6¶
The definitive open-source knowledge base for SRv6.
Learn how SRv6 works. Build labs. Contribute.
- Leaf–spine fabric Spines, leaves and edge PEs, every link IPv6
- IS-IS underlay Each node advertises its uSID locator · RFC 9352
- Flex-Algo 128 Metric: link delay. The low-latency plane, with its own locators · RFC 9350
- BGP services PE-B advertises a VPN SID (End.DT46) · RFC 9252
- AI fabric · MRC GPU NICs spray over static uSID paths, no BGP or IS-IS
Learn SRv6: a free course¶
Learn SRv6 in order, from your first SID to TI-LFA, L3VPN, migration from MPLS, vendor configuration and AI fabrics. Learn SRv6 organizes every guide on this site into tracks, with hands-on labs, and video lessons coming to YouTube.
SRv6 for AI¶
SRv6 now runs inside some of the largest AI training networks in the world. OpenAI, Microsoft and Oracle use static SRv6 uSID source routing in their GPU fabrics through MRC, with no BGP in the backend. Meta and Alibaba use SRv6 traffic engineering to connect AI clusters across sites. The SRv6 for AI section covers who runs it, how it works, how to design a fabric, and the standards behind it.
SRv6 for AI: the complete guide How MRC uses SRv6
SRv6 SID Decoder¶
Paste any SRv6 SID or uSID container and see what every bit means: Locator-Block, node, function and argument, F3216 uSIDs, the end-of-container, and a hop-by-hop walk of the RFC 9800 shift-and-forward. It also looks up any IANA behavior codepoint, and every decode has a link you can share.
State of SRv6 2026¶
How far has SRv6 come? Our data report counts 23 SRv6 RFCs, 65 active IETF drafts, 19 organizations with public deployments and uSID support across 11 implementations, with a source for every number and charts you can reuse.
Read the State of SRv6 2026 report
SRv6 Documentation and White Papers¶
The SRv6 documentation index links the official SRv6 docs of every vendor and open-source project, from Cisco, Juniper and Nokia configuration guides to the Linux kernel, FRRouting and SONiC. The SRv6 white paper library collects every public SRv6 white paper with a summary, and the free SRv6 White Paper 2026 covers architecture, design and migration from MPLS.
SRv6 documentation SRv6 white paper (PDF)
Explore¶
Topics¶
New to SRv6? Start here. Understand the core concepts of Segment Routing over IPv6, SID structure, and network programming.
Use Cases¶
Discover real-world applications: Traffic Engineering, VPN services, service function chaining, and more.
Implementations¶
Hands-on guides for Linux, Cisco IOS-XR and IOS-XE, Juniper, Arista, Nokia, FRRouting, and SONiC.
Labs¶
Build your own SRv6 lab with netlab, Containerlab, and GNS3. Copy-paste configs included.
RFCs & Standards¶
Quick reference for all SRv6-related RFCs and IETF drafts with plain-English summaries.
Community¶
Join the community! Learn how to contribute, report issues, or suggest new content.
SRv6 (IPv6 Segment Routing) in a Nutshell¶
SRv6 (Segment Routing over IPv6) is a networking technology that uses IPv6 itself to steer and program traffic. The source router writes the path into the packet as an ordered list of IPv6 addresses called SIDs, and the network follows it without MPLS, LDP or RSVP-TE. The complete guide is in What is SRv6?.
It is also written Segment Routing v6, SR v6 or IPv6 Segment Routing. Technically, SRv6 is the IPv6 data plane of the Segment Routing architecture (RFC 8402). It encodes forwarding instructions directly in IPv6 packet headers. Instead of relying on MPLS labels and signaling protocols like LDP or RSVP-TE, SRv6 uses native 128-bit IPv6 addresses as Segment Identifiers (SIDs) to program network paths and behaviors.
The source node builds an ordered list of segments that defines the packet's path and the processing at each hop. The active segment sits in the IPv6 destination address; when there is more than one, the rest travel in a Segment Routing Header (SRH) — and with uSID compression a whole path often fits in the destination address with no SRH at all. Transit routers simply forward on the destination address. This enables traffic engineering, VPN services, fast reroute and service function chaining on a plain IPv6 network.
Why choose SRv6 over traditional MPLS?
- Simplified operations — No more MPLS label management, LDP, or RSVP
- Native IPv6 — Uses the existing IPv6 data plane with no overlays
- Network programmability — Define custom network behaviors with SRv6 functions (End, End.DT4, End.DX6, etc.)
- Scalability — 128-bit SID space supports massive networks and micro-segmentation
- Unified data plane — Single protocol stack for transport, VPN, and TE services
The trade-offs, with real header overhead numbers, are in SRv6 vs MPLS.
What is SRv6? The full explanation See how SRv6 works SRv6 glossary
This is a community project
SRv6.md is built and maintained by the networking community. Every page has an edit button () — if you spot an error or want to add content, we welcome your contributions!
SRv6.md is not affiliated with any vendor. Content is provided under the MIT License. About SRv6.md and how content is verified.