IP Networks within Data Center and Transport Connectivity
This domain covers data-center fabrics and transport connectivity: leaf-spine underlays, EVPN-VXLAN overlays, ECMP, anycast gateways, multihoming, SR Linux, and transport technologies such as segment routing and optical transport. Questions often test how an overlay depends on a stable underlay.
Core rules
- The underlay is usually Layer 3 with OSPF, IS-IS, or BGP and provides loop-free ECMP paths between leafs.
- EVPN is the control plane; VXLAN is the data plane. VNIs are 24-bit and map to bridge domains.
- EVPN route types include type 1 Ethernet auto-discovery, type 2 MAC/IP, type 3 inclusive multicast, type 4 Ethernet segment, and type 5 IP prefix.
- Multihoming uses Ethernet Segment Identifiers (ESI) so a host can dual-home to leafs without STP.
- Transport may use SR-MPLS or SRv6 for segment routing and WDM/OTN for optical capacity.
Worked example
A VM on leaf1 in VLAN 100, VNI 10100, sends to a VM on leaf2. Leaf1 learns the destination MAC through an EVPN type 2 route and encapsulates the frame in VXLAN with VNI 10100. The underlay ECMP-forwards the packet to leaf2’s VTEP, which decapsulates it and forwards on VLAN 100. If leaf2 expects VNI 10200, the packet arrives but lands in the wrong bridge domain. The underlay MTU must also allow for VXLAN overhead.
Traps that cost marks
- Mixing VLAN and VNI mappings between leafs.
- Forgetting EVPN route type meanings.
- Ignoring underlay MTU and ECMP hash behavior.
- Assuming an overlay fixes a broken underlay.
Key points
- Underlay provides reachability; overlay provides services.
- EVPN controls VXLAN MAC/IP learning.
- Type 2 carries MAC/IP; type 3 handles multicast.
- ESI enables EVPN multihoming.
- MTU must account for encapsulation overhead.