Deployment Guide

11 Leaf-Spine Deployment and Best Practices Guide | Version 1.0
3.2 Layer 2 leaf-spine topology
In a layer 2 leaf-spine network, traffic between leafs and spines is switched (except for a pair of edge leafs) as
shown in Figure 8. VLT is used for multipathing and load balancing traffic across the layer 2 leaf-spine fabric.
Connections from hosts to leaf switches are also layer 2.
For connections to external networks, layer 3 links are added between the spines and a pair of edge leafs.
Rack 2
Rack 1
Leaf 4
Leaf 3
VLTi
Leaf 2
Leaf 1
VLTi
Rack n
Edge Leaf
Edge Leaf
VLTi
Spine 1 Spine 2Spine 1 Spine 2
L3 Connection
L2 Connection
VLTi
VLT
Host
VLT
Host
VLT
Host
VLT
External
Network
Layer 2 leaf-spine network
3.3 Design considerations
When compared to a layer 3 topology, a layer 2 topology is generally less complex but has some limitations
that must be considered. These include:
• For each VLAN, the layer 2 topology creates one large broadcast domain across the fabric. The layer
3 topology has the benefit of containing broadcast domains to each rack.
• The layer 2 topology is limited to 4094 VLANs across the fabric. The layer 3 topology allows up to
4094 VLANs per rack.
• The layer 2 topology is limited to two physical switches at the spine layer (configured as VLT peers).
In a layer 3 topology, additional spines may be added as needed to provide additional paths and
bandwidth. Therefore, a layer 3 topology is more scalable and is better suited for very large networks.
• Overlay networks utilizing VXLAN (such as VMware NSX) require a layer 3 underlay network.
If none of the layer 2 limitations are a concern, it may ultimately come down to a matter of preference. This
guide provides examples of both topologies.