A spine and leaf network topology (also known as leaf-spine network) is a two-tier design that is intended to overcome the limitations of the traditional three-tier network architecture and provide improved performance.

The traditional three-tier network topology includes access switches, aggregation routers (sometimes referred to as the distribution layer), and core routers. The access layer switches sit closest to end users and feed into the aggregation layer which, in turn, feed into the core layer routers.

Network architects configure the network with redundant pathways to ensure traffic can always get around any potential network failure, typically employing the Spanning Tree Protocol (STP). STP always has two potential routes for traffic to take, but deactivates all except the primary route.  Should the primary path fail, STP will bring up a backup path which it uses until the primary comes back online.

image of spine and leaf network topology

While STP works in networks that are not experiencing heavy traffic loads, it can result in backups under heavy loads, such as in a data center.

The leaf spine topology was developed to overcome this limitation.

Originally implemented in data center networks, the topology has just two layers: leaf and spine switches. Leaf switches constitute the access layer, connecting to data center end devices such as servers, firewalls, edge routers and load balancers. Each leaf switch connects to multiple spine layer switches, for redundancy.

The leaf spine topology works well because it’s a good fit for the “east-west” nature of data center traffic, meaning traffic that flows predominantly inside the data center – such as from server to server – rather than outside, in a client-server (or north-south) fashion.

Imagine a rack full of servers in a data center. At the top of the rack will be a pair of leaf switches, also appropriately known as Top of Rack (ToR) switches. For redundancy, each server in the rack connects to both of the ToR switches.

Each of the ToR leaf switches then connects to multiple spine switches; in a data center configuration, it’s likely that each leaf switch connects to every spine switch. In such a topology, there’s no need for spine switches to connect to each other, because every leaf switch is able to connect to every spine switch.

Additionally, this topology enables all ports on a spine switch to be used to connect to leaf switches, using either Layer 2 switching or Layer 3 routing.

From a logical perspective, all switches are the same distance from one another, just a single “hop.” That creates a dramatic performance improvement vs. the three-tier topology using STP.

While the spine and leaf network topology is clearly a benefit for data center networks, what about the rest of the enterprise network?

Companies are starting to address the enterprise issue. Pica8, for example, is a white box network operating system (NOS) vendor whose PICOS technology extends the spine and leaf topology throughout the enterprise.

PICOS eliminates much of the complexity of traditional three-layer network topologies by enabling hundreds of leaf and spine switches to be managed as though they were a single logical switch, with a single IP address.

The PICOS white box NOS retains all the Layer 2/Layer 3 features and functions of the traditional three-tier design, including a resilient, multi-path fabric architecture that ensures high availability. It also supports software-defined network (SDN) protocols and programmability. And it’s a highly scalable architecture, based on cost-effective white box networking infrastructure.