Fat-Tree Topology
Back to Glossary
Networking
Spine-Leaf
A two-tier network topology in which every leaf switch connects to every spine switch, providing predictable low-latency east-west traffic paths — the dominant architecture for modern datacentre fabrics.
Full Definition
The spine-leaf topology replaces the traditional three-tier (core-aggregation-access) design in modern datacentres. The fabric has two layers: leaf switches, which connect servers, storage, and endpoints; and spine switches, which interconnect all leaf switches. Every leaf connects to every spine via high-speed uplinks, creating a non-blocking full-mesh inter-switch fabric.
This design delivers any-to-any east-west traffic with predictable latency — typically 2–3 hops between any two endpoints — which is critical for distributed computing, AI training, and storage fabrics. Scaling is horizontal: adding servers means adding a leaf switch; adding leaf switches means adding spine switches. Modern spine-leaf fabrics use 100 GbE or 400 GbE uplinks with BGP EVPN/VXLAN overlay networking. While mathematically equivalent to fat-tree, spine-leaf specifically describes the physical and operational implementation pattern.
Also Known As
leaf spineClos networkspine leaf topology
Source Reference
Arista Networks Datacenter Design Guide; Cisco Validated Design for Spine-Leaf Architecture