What is the spine-leaf network architecture?

Traditionally, data centers would be structured using a three-layered architecture like the one illustrated below.

However, this model is no longer ideal due to profound changes in data centers over the last decade, particularly regarding the direction of data flow. Where this model focused on North-SouthNorth-South traffic refers to a data flow that enters/leaves the data center to/from a system physically residing outside the data center. traffic, a new model known as the spine-leaf architecture focuses on East-WestEast-west traffic refers to data flow from server to server within a data center. traffic instead.

Description

Designed to address the shortcomings posed by the three-tiered model, the surge of cloud and containerized infrastructure, and the increase in East-West traffic, the spine-leaf architecture is deployed extensively in modern data centers.

As its name suggests, the spine-leaf architecture is a two-tiered network architecture composed of a leaf layer and a spine layer.

Leaf layer

The leaf layer contains leaf switches, known as top-of-rack (ToR) switches that are connected to network and storage devices.

Spine layer

The spine layer acts as the network's "backbone" and contains spine switches that connect all leaf switches to each other. Every spine switch is connected to every leaf switch, and vice versa. However, spine switches are not connected to each other.

A spine-leaf architecture looks like the following:

Advantages

Due to its structure, the spine-leaf architecture comes with a host of advantages, namely:

• Lower latencyLatency is the term used to describe delays in communication over a network.: To send a packet from one device to another in a three-tiered network, the packet has to travel through two distribution layer switches and one core layer switch. This is a total of 3 switches. However, in a spine-leaf model, the packet only has to travel through one spine layer switch.

• Minimized traffic bottlenecks: The reduced latency leads to a more efficient data transfer and minimizes traffic bottlenecks.

• Large non-blocking fabricA network fabric is the mesh of connections between network devices.: Since every leaf layer switch connects to every spine layer switch, with neither being interconnected, a non-blocking fabric is made.

• Increased scalability: If an oversubscription of a particular spine layer switch occurs, a new spine layer switch can be introduced and linked to all of the leaf layer switches. If a leaf layer switch's ports are being overused, a new leaf layer switch can be added with connections to all the spine layer switches.