AI Datacenter SRv6 Loadbalancing Transport

Stand up the IPv6 dual-stack underlay that the SRv6 transport will ride on — additive alongside the IPv4 underlay and EVPN-VXLAN overlay you already built. With the /127 link addresses pre-set, configure a second numbered IPv6 BGP session on every switch — the spines and the leaves — advertise each leaf's uSID locator /48, and confirm every locator is reachable via both spines with per-flow ECMP. No SRv6 dataplane yet — that's the next lab.

• Dual-stacking a live fabric — a second, numbered IPv6 BGP session added alongside the IPv4 / VXLAN you built, with zero disruption
• Why a separate numbered IPv6 session (not an unnumbered interface peer) is what coexists with the existing IPv4 underlay on each link
• Advertising an aggregate uSID locator /48 with a blackhole anchor + no bgp network import-check
• maximum-paths and reading IPv6 ECMP nexthops as link-locals (fe80::%ethN)
• Why per-flow ECMP across both spines is the foundation any AI-fabric transport depends on

Starting from the unnumbered IPv6 underlay + ECMP you built in Lab 5, lay a Segment Routing over IPv6 (SRv6) uSID transport across the fabric. Define micro-SID locators in FRR, program kernel End.DT6 endpoints so each leaf decapsulates, steer flows into uSID with an H.Encaps headend — then watch distinct flows spread per-flow across both spines on the IPv6 flow label, the SRv6 analog of the VXLAN ECMP you saw in Lab 4.

• SRv6 and the uSID (micro-segment) compressed SID format — block / node / function bits
• FRR SRv6 locators as the control-plane object (behavior usid)
• Kernel End.DT6 endpoints — and why seg6local must bind to a real device, never lo
• H.Encaps.Red headend — steering a flow into a single uSID with no SRH overhead
• Per-flow ECMP via the IPv6 flow label — how SRv6 load-spreads collective traffic across spines
• SRv6 vs VXLAN as an AI-fabric transport — what each buys you