The world of IT is continuously changing, sometimes even daily. My last blog post was written with the intent of introducing a new type of network hardware configuration model. This is exactly what it turned out to be – only with a plethora of technical jargon meant to illustrate the complex configurations and convey the ideals behind this new architecture. Now, I want to introduce this complex idea from the dynamic world of enterprise networking to individuals without a technical background.
What does a Leaf-Spine network look like?
A Leaf-Spine network is comprised of two layers of networking equipment. The backbone of the network is the Spine layer, while secondary activity occurs on the Leaf layer. Every router (a device used to link different networks together) on the Spine layer connects to every switch (or point of connection between devices on a network) on the Leaf layer. With this network design, every Leaf switch on the network is the same distance away. The switches on the Leaf layer connect to every other device on the network, including: servers that provide a service, load balancers that spread out traffic across links or to different devices, or edge routers that connect to someone else’s network or the common internet.
Figure 1: Leaf-Spine network design
Why did this new type of model appear? What happened with the old model?
Leaf-Spine was developed to utilize all available network links between devices at the same time. With the older, three-tier model, there were multiple links between devices on a network. However, because of the underlying protocols needed to move the traffic and keep the design accurate, the redundant paths were disabled until the active path was disconnected or otherwise failed. Leaf-Spine, using a variety of networking protocols, keeps all links active and removes any chance for a loop to be created.
Users’ desire to seamlessly add hardware to existing networks, create more connections to devices, and increase the data capacity of the network also encouraged the development of Leaf-Spine network architecture. A new Leaf or Spine device can be procured, cabled accordingly and set up with the configuration needed to talk to devices on the existing network; the new device will then be moving traffic. When procuring and setting up additional hardware on the legacy, three-tiered network architecture, the configuration aspect was demanding for the IT department. Device configurations had to be hard-coded into devices; Leaf-Spine hardware configurations can be dynamically created, assisted by software. These advantages to Leaf-Spine architecture are unique to this new type of hardware configuration and setup.
There are a few limitations of the Leaf-Spine design. The first is the amount of planning and sheer length of cable needed to connect every Spine router with every Leaf switch. As the physical layout of the design changes, planning for and gathering the cables becomes a hassle.
The other concern is that using the specific protocols restrict other useful, mainstream networking tools. Separate virtual networks, or Virtual Local Area Networks (VLANs), do not work on the physical layer of the Leaf-Spine network. There is a tool called Software Defined Networking (SDN) that places VLANs and other tools back into the IT department’s hands by layering a virtual network on top of the existing Leaf-Spine network.
Why would a business want to implement these changes?
The idea for a Leaf-Spine network starts with the ideals of being fast, predictable, and scalable. A network is responsible for transferring data and information; it can be from a user’s computer to another computer, to the internet, or from an application stored in the company data center. The best networks do this as quickly as possible.
Another desirable characteristic of a network is predictability. For data travel to be predictable, the network and its design must have information moving in a clear, well-defined path from one device to another. Lastly, for a network to be scalable, the devices must know exactly how to handle and negotiate changes in the design.
Leaf-Spine accomplishes these goals with a two-tiered hardware architecture, new network protocols, and device configurations that allow for utilizing all links and negotiating dynamic changes when needed. Leaf-Spine is not the perfect solution for all networks, but it offers significant advantages if planned for and implemented diligently.