Cisco FabricPath Spine‑and‑Leaf Network Architecture Overview

Cisco introduced FabricPath technology in 2010, offering new capabilities and design options that enable network operators to build Ethernet fabrics with higher bandwidth availability, design flexibility, and reduced deployment and operational costs.

Encapsulation Format and Standards Compliance

FabricPath spine‑and‑leaf networks are Cisco proprietary but based on the TRILL standard, using FabricPath‑MAC for MAC‑frame encapsulation.

Underlying Network

The fabric uses a Layer‑2 FabricPath MAC encapsulation and FabricPath IS‑IS as the control plane; each switch is identified by a FabricPath switch ID, and the IS‑IS control plane builds reachability information.

Overlay Network

FabricPath does not have a separate overlay control plane; host information is learned via flood‑and‑learn mechanisms.

Broadcast and Unknown Unicast Traffic

FabricPath IS‑IS creates two multicast trees that carry broadcast, unknown unicast, and multicast traffic across the fabric, flooding these frames to all edge ports within a VLAN or broadcast domain.

Host Detection and Reachability

Switches rely on initial data‑plane traffic flooding to learn host reachability, which can become problematic as the number of hosts grows; multi‑topology features help limit flooding within sub‑areas.

Multicast Traffic

Layer‑2 multicast is forwarded via the multicast trees, while Layer‑3 IP multicast uses Protocol‑Independent Multicast (PIM) after routing to the target VLAN.

Layer‑3 Routing Function

FabricPath supports Layer‑3 routing on spine or border leaf switches via SVIs, providing internal and external routing with up to four anycast gateways; design choices include internal/external routing at border spines or border leaves.

Design Example: Internal/External Routing at Border Spine

Spine switches act as both Layer‑2 and Layer‑3 boundaries, handling VLAN‑internal FabricPath frame forwarding and inter‑VLAN routing via SVIs, with default gateways reachable in a single hop.

When more than four spines are present, the control plane and MAC learning must be distributed, and MAC address scalability limits must be considered.

Design Example: Internal/External Routing at Border Leaf

Border leaf switches perform Layer‑2 FabricPath forwarding only, while SVIs on these leaves handle inter‑VLAN routing and exchange routing adjacency for external traffic; routing traverses leaf‑to‑spine then to the destination leaf.

Multitenancy

FabricPath supports Layer‑2 multitenancy via VN‑segment (VXLAN‑like) using a 24‑bit namespace, allowing VLAN reuse across leaves and providing isolation; up to 16 000 000 VN segments are theoretically possible.

Layer‑3 multitenancy is achieved with VRF‑lite, supporting up to 4096 VLANs per VRF.

Cisco FabricPath Spine‑and‑Leaf Network Summary

FabricPath provides a simple, flexible, stable, and scalable Ethernet fabric with fast convergence, leveraging flood‑and‑learn for host discovery and IS‑IS for control; it supports up to four anycast gateways, centralized Layer‑3 routing, and both Layer‑2 and Layer‑3 multitenancy.

Feature Table

Item

Description

Transport medium

Layer 1

Encapsulation

FabricPath (MAC‑in‑MAC frame encapsulation)

Unique node identifier

FabricPath switch ID

End‑host detection

Flood and learn

Silent host discovery

Yes

End‑host reachability and distribution

Flood and learn plus conversational learning

Broadcast and unknown unicast traffic

Flood by FabricPath IS‑IS multidestination tree

Underlay control plane

FabricPath IS‑IS

Overlay control plane

–

Layer 3 routing function

Internal and external routing at border spine; internal and external routing at border leaf; up to 4 anycast gateways supported

Multicast traffic

Layer 2 via multidestination tree; Layer 3 IP multicast via PIM

Multitenancy

Layer 2 with VN‑segment; Layer 3 with VRF‑lite

Standard reference

TRILL based (Cisco proprietary)

Supported hardware

Cisco Nexus 7000 series (including 7700), Nexus 5500/5600, Nexus 6000 series