We often use broad descriptions when describing network topologies. In this video, you’ll learn about star, mesh, hybrid, spine and leaf, and other network topology types.
Networks can connect to each other in many different ways. And in this video, we’re going to look at a number of different network topologies and how those different network topologies can be deployed in an enterprise network. This can be very useful during the planning process, where you’re designing the network. It’s also useful during the troubleshooting process so that you can tell how different networks are connected to other networks. This is also useful during the troubleshooting process so that you can visually see how one network might connect to and send data to another network.
Let’s start this conversation with one of the most popular network topologies in use. That would be a star network. Sometimes you may hear this referred to as a hub-and-spoke network. This is what we use in most large networks to connect devices together. There’s a central device that is used as the central networking component. And then everyone on the network is all connecting to that same central device.
A good example of a star network, or a hub-and-spoke network, is one that we have with switched Ethernet. Our Ethernet switch sits in the middle of the network. And all of the other devices connect to that same Ethernet switch. If each of these spokes around the network would like to communicate with each other, they all must communicate through this central hub.
A mesh network design is when one location or device connects to another location or device over more than one network connection. This means that we can have one link communicating to another. And it might follow one path to get to that location. Or it may follow an alternate path to be able to get to that location.
One reason we design a network this way is so that if one link in the mesh was to fail, we can use one of the other links to complete that communication. We might also perform load balancing over these links, where half of the data is sent over one connection, and the other half of the data is sent over a different connection. Although we can certainly design mesh networks for local area networks, we often see them deployed in wide area networks so that we can maintain connectivity to a remote site regardless of what network connection might be available.
If you were to look at a large enterprise network, you would notice there are a number of different architectures all being used in different parts of the network. And when we combine all of those together, we’ve created a hybrid network. One part of the network might be a star network. Another part of the network may be point to point. And a third part of the network might be a mesh network. Once we combine all of these together, we’ve created a hybrid architecture.
Many data centers take advantage of a spine and leaf architecture. This is where you would have individual switches at the top that are the spine of the network. There would be switches in the middle that are leaf networks. And then you would have different devices connect to the leaf.
You’ll notice that the spine network is connected to all of the leafs. And the leafs are connected to each of the spines. However, those leaf switches do not connect directly to each other. And the spine switches also do not connect directly to each other. This works very well for an architecture we use in many data centers, known as top-of-rack switching. In top-of-rack switching, every rack has a leaf switch at the top of every physical rack. We would then connect all of the physical devices in that rack to the leaf switch that is at the top of that rack.
This keeps your cabling very simple because all of your cabling is self-contained within that same rack. There is redundancy because you have this rack connected to multiple spines. This also increases the overall performance of the network because we’re no more than one switch away from any other device in the data center.
If your data center only has a handful of racks, this may be a relatively inexpensive way to connect all of these devices together. But when you have tens or hundreds or even thousands of racks in a data center, you would need a separate switch for every rack. And of course, the cost of that will increase for each rack you have in the network.
When looking around the network, you may find some wide area network connections that are point to point. And as the name implies, there is a single point connected to a single point. This was a very common design on older wide area networks, where you would use something like a T1 or a T3 connection. We refer to those as point-to-point T1 or point-to-point T3. We might also use this design in local area networks. If you work on a campus, you can connect one building to another over two connections. And the connection between those two buildings would be a point-to-point connection.