SRv6 Workshop · Netdev 0x1A · Rome 2026

SRv6 in Linux: AI backends, Provider Deployments,
L2 services
and Eco-systems updates

Third edition of the netdev SRv6 workshop, after Lisbon 0x16 (2022) and Zagreb 0x19 (2025).

Chair: Stefano Salsano — University of Rome Tor Vergata / CNIT
Netdev 0x1A — The Technical Conference on Linux Networking (13–16 July 2026, Rome, Italy)
Workshop: Monday 13 July 2026, 10:00–12:00

Programme & slides 10:00–12:00

TimeTalkSpeakerSlides
10:00–10:05 Workshop Introduction Stefano Salsano
Univ. of Rome Tor Vergata
10:05–10:15 SRv6 Introduction Ahmed Abdelsalam
Cisco
10:15–10:25 SRv6 for the AI Backend Deterministic path placement, decided at the host and the NIC Stefano Salsano
Univ. of Rome Tor Vergata
10:25–10:40 SRv6 in SONiC “Enabling AI Backend use-case and many others” Ahmed Abdelsalam
Cisco
10:40–10:55 SRv6 in FRR “5 Years of mainline support” Carmine Scarpitta
Cisco
10:55–11:15 SRv6 for network providers and service protection Provider-style design on Linux + FRR, observability, redundancy protection Ferenc Fejes
Ericsson
11:15–11:25 SRv6 in Linux “9 Years of mainline support” Andrea Mayer
Univ. of Rome Tor Vergata
11:25–11:45 SRv6 Layer 2 Services support End.DT2U, the sr6 Ethernet pseudowire device, and the road past VXLAN Andrea Mayer
Univ. of Rome Tor Vergata
11:45–12:00 Recap and Discussion all speakers

Slides are self-contained Reveal.js decks: they run in any browser, also offline. Navigate with the arrow keys or a presenter remote; press F for fullscreen. Decks marked soon will be posted after the session.

The workshop

Segment Routing over IPv6 (SRv6, RFC 8986) lets an application or operator encode a packet-processing program directly in the IPv6 header. SRv6 has been supported in the Linux kernel since release 4.10, and a rich open-source ecosystem (FRR, SONiC, Cilium, VPP) has grown on top of it.

Following the SRv6 workshops at Netdev 0x16 (Lisbon 2022) and 0x19 (Zagreb 2025), this edition covers three directions that directly impact the kernel datapath — source-routed AI backends, L2 services beyond VXLAN, and provider-grade SRv6 deployment with service protection — plus a short ecosystem update.

Three directions

01

Source-routed AI backends

ECMP spreading gives little control over how flows map onto a fabric you own completely. SRv6 moves path placement to the host and the NIC: deterministic, congestion-aware, and reacting to failures without waiting for control-plane convergence. In Linux: per-packet segment-list selection in seg6/seg6local, eBPF/XDP and NIC offload at line rate. Underpins the industry MRC protocol (main-track talk).

02

L2 services beyond VXLAN

RFC 8986 defines the L2 endpoint behaviors (End.DX2, End.DT2U, End.DT2M) and RFC 9252 the EVPN overlay over SRv6. Kernel L2 support still stops at the End.DX2 cross-connect. A recent netdev RFC series adds End.DT2U and the sr6 Ethernet pseudowire device — the path to multipoint services, EVPN-over-SRv6 in FRR, and replacing VXLAN in cloud orchestrators.

03

Providers & service protection

End-to-end provider-style design on GNU/Linux + FRRouting: addressing, per-service VRF, SID allocation, BGP/IS-IS, TE, with an SR-MPLS coexistence comparison. Lookup-level observability with tablesnoop, and Redundancy Protection (R-SID, draft-ietf-spring-sr-redundancy-protection, XDPFRER for SRv6).

References

Notes

Workshop chair: Stefano Salsano, University of Rome Tor Vergata / CNIT. Speakers: Ahmed Abdelsalam (Cisco), Carmine Scarpitta (Cisco), Ferenc Fejes (Ericsson), Andrea Mayer (Univ. of Rome Tor Vergata).

Contact: stefano.salsano@uniroma2.it