BookMyFabricBookMyFabric

Campus EVPN-VXLAN Pod — Overview

A compact enterprise pod that builds the same EVPN-VXLAN overlay as the AI fabric, at campus scale: 2 spines, 2 leaves, 2 PCs.

This course runs on its own dedicated topology — selecting it stands up the campus pod (and tears down any other fabric). Distinct device names make it obvious which fabric is live:

RoleDevices
Spinescspine1, cspine2
Leavescleaf1, cleaf2
Hostscpc1 (on cleaf1), cpc2 (on cleaf2)

Addressing — a 2x2 CLOS underlay (10.1.x.x/31 links, 10.0.x.x loopbacks) with a stretched overlay segment VNI 10100 / 192.168.100.0/24. cpc1 is 192.168.100.11, cpc2 is 192.168.100.13.

Goal: the two PCs sit on different leaves but the same overlay segment — so cpc1 can ping cpc2 over EVPN-VXLAN, no L3 hops between them.

Run this lab hands-on — free

Everything below runs on a live SONiC/FRR fabric you drive from the browser. Book a free fabric slot, start the Campus SONiC EVPN-VXLAN Fabric course, and work each step with real consoles and graded checkpoints.

Campus EVPN-VXLAN Pod — Walkthrough

This pod boots canonical — the underlay and EVPN-VXLAN overlay are already configured. Your job is to confirm it works end to end.

Step 1 — Underlay BGP

Open a console on cleaf1 and check the fabric came up:

show bgp summary

You should see 2 Established sessions (to cspine1 and cspine2).

💡 Why this matters — the overlay rides on the underlay. No underlay BGP, no VTEP reachability, no overlay.

Step 2 — Overlay reachability

cpc1 and cpc2 are on different leaves but the same stretched segment (192.168.100.0/24, VNI 10100). From cpc1:

ping 192.168.100.13

It should succeed — the frames ride an EVPN-VXLAN tunnel between cleaf1 and cleaf2, with no L3 hop in between.

💡 Why this matters in AI DCs — this is the exact same overlay primitive the AI fabric uses for the GPU plane, just at 2-leaf campus scale.

For teams

Training a network team on this?

We run this curriculum for enterprises — private per-learner fabrics, EVPN-VXLAN, SRv6 and telemetry, self-paced or instructor-led.

Team training