Do you know spine switch?

What is a spine switch?

The data center’s Ethernet switching design has undergone major changes over the past decade and now includes a backbone switch as part of the data center architecture. In recent years, architectures called leaves and spine or distributed cores have become the leading design in data centers. This design requires a spine switch specifically designed for the spine to work.

The traditional network structure model includes three layers of switches: core, aggregation, and access. This model is sufficient for basic network setup and vertical traffic. However, as data center networks grow, faster, more efficient horizontal or “something” traffic is needed. In addition, many data centers use spanning tree protocols to prevent network loops but can block up to 50% of switch ports and limit broadband usage. This is where the leaves and spine structures become popular.

Spine Switch: The core part of the network structure

All backbone switches can handle Layer 3 (L3) with high port density for scalability. In Software Defined Networking (SDN), the backbone switch is directly connected to the network control system with a virtual Layer 2 switch at the top of the leaf backbone system. The controller is not part of the data path but is still the central network engine, also known as the SDN controller.

Each L3 port on the backbone switch is dedicated to the underlying (L2) leaf switch. Servers, access points or firewalls cannot be connected or found on the backbone switch, they are only used to connect to the leaf switch. One connection method is to use equal cost multipath (ECMP). In this architecture, each leaf switch is connected to each trunk switch. Therefore, the number of uplink egresses on a leaf switch determines the number of backbone switches that can be in the network. This model allows each connection to have exactly the same travel distance without any difference in the structure.

Benefits of the spine leaf system

Some of the benefits of this system involve some predictability and latency. With the leaf ridge architecture, you always know the amount of docking that each packet must pass. In the old architecture, the payload from one server might have to go through multiple switches to communicate with different servers on the same network. Therefore, the traffic and distance it must cover are consistent, which also results in fewer bottlenecks between access layer switches. In addition, the removal of STP to support this level 3 routing allows for a more stable environment.

Spinal switch

Another benefit of this structure is scalability. When the traffic on an active link is too large, adding additional spines increases the inter-layer bandwidth and reduces the amount of “traffic congestion” in the network. Adding new spikes or leaves to the system will not destroy the L2 protocol.

This method can also be assisted by using network coverage (eg VXLAN). These are most useful in high-capacity, multi-tenant environments in Infrastructure as a Service (IaaS).

Original article from: https://www.sdxcentral.com/sdn/definitions/what-are-spine-switches/