The Layer 3 Leaf and Spine design is becoming the more adopted design for newer, more modern data center fabrics depicted in the following figure:
Figure 33. Leaf and Spine Layer 3 Design
It has the following characteristics:
- L3 is terminated at the leaf, thus all the VLANs originating from ESXi hosts terminate on leaf.
- The same VLANs can be reused for each rack.
- The leaf switches provide default gateway functionality.
- The NSX-T T0 Gateway for the WLD will peer with the leaf switches in one rack.
- Vendor agnostic - Multiple network hardware vendors can be used in the design.
- Reduced VLAN span across racks, thus smaller broadcast domains.
- East/West for an NSX domain can be confined within a rack with intra-rack routing at the leaf.
- East/West across NSX domains or cross-rack is routed through the spine.
- NSX-T Tier0 peering is simplified by peering the WLDs with the leaf switches in the rack.
- The Layer 2 VLANs cannot span racks. Clusters that span racks will require a solution to allow VxRail system traffic to span racks using hardware VTEPs.
- The Layer 3 configuration might be more complex to implement.