Spine and leaf architecture - reducing latency

A spine and leaf architecture, ensures that communication between any two entities on any two leaf switches will have only two physical links to traverse, and never more than two. This optimizes latency. If a data path went from spine to spine as well, you may have three or more such links to traverse.

The following notes explain which connections are required for spine and leaf switches:

Spine and leaf architecture - connections to spine switches Spine and leaf architecture - connections to leaf switches