Which statement is correct about a three-stage IP fabric underlay?

In a three-stage (leaf-spine-leaf) IP fabric, which is a type of Clos network, traffic always travels from a source leaf to a spine, and then to a destination leaf. This architecture ensures a predictable, non-blocking, and low-latency path. Any server connected to an ingress leaf is always one hop from a spine and a second hop from any other egress leaf in the fabric. This two-hop intra-fabric path is a key design principle for scalability and performance.

B. Spine devices do not connect to other spine devices in a standard Clos architecture; they only connect to leaf devices.

C. Leaf devices do not connect to other leaf devices directly; all east-west traffic must traverse a spine device.

D. While multihoming servers to multiple leaf devices is a common best practice for redundancy, it is not a mandatory requirement of the fabric's architecture.

1. Juniper Networks. (2017). Data Center Fabric Architecture Design Guide. "In a Clos network

every lower-tier switch (leaf) is connected to each of the top-tier switches (spine) in a full-mesh topology... This topology provides high bandwidth between the leaf switches because traffic can be spread across all the spine switches. The path from one leaf switch to another is always two hops..." (Page 8).

2. Juniper Networks. (n.d.). IP Fabric Underlay Network Design. Junos OS Documentation. "In a spine-and-leaf architecture

also known as a Clos network

each leaf device connects to each spine device in the network. This creates a large

non-blocking fabric that provides high-speed connectivity with predictable

low latency and is highly scalable."