Designing of Spine-Leaf Architecture
To design spine-leaf architecture, the following are some considerations:
- Oversubscription rate: It is the rate when all the devices are sending data (traffic) at the same time. It can be measured in the traffic entering or leaving the data center and in traffic between the devices in the data center. It is the ratio between the bandwidth of backbone switches and to the capacity of servers.
- Leaf and spine sizing: The ports in the spine switches determine the maximum number of leaf switches in the topology. The number of spine switches is governed by the combination of the required throughput between leaf switches, the number of redundant paths, and their port density. So, the number of spines and leaf is to be considered to avoid network issues.
- Layer 2 and Layer 3 design: Spine-Leaf architecture is made of two layers which can be built at either Layer 2 or Layer 3. Layer 2 design provides flexibility, and allows VLAN span and MAC address to migrate anywhere. Layer 3 design provides the fastest convergence times and maximum scale ECMP fan-out supporting up to 32 or more spine switches.
Spine-Leaf Architecture
Spine-Leaf Architecture is a two-layer architecture. This architecture has two layers i.e The spine layer and the Leaf layer. The Spine layer and Leaf layer are connected in the full mesh topology. This architecture overcomes the limitations of three-layer hierarchical architecture. Every leaf switch is connected to a spine switch in a full mesh topology. The leaf layer contains the access switches that connects to the servers and provide a connection for end users. The spine layer is called the backbone of the network as it interconnects all the leaf switches.