A foundational knowledge of network architecture starts with the concepts of broadcast domains and collision domains. In this video, you’ll learn how modern network design has eliminated collision domains and how you can minimize the impact of broadcasts through the use of properly engineered broadcast domains.
When half duplex ethernet networks were all the rage, we had to be very concerned about collision domains. These days, it’s hard to find a collision domain because we use full duplex ethernet. And the term collision is a bit misleading. If you’re driving a car, a collision is something you never want to occur. But on a half duplex ethernet network, a collision is simply the normal process that occurs when people are trying to communicate on the network and it’s perfectly expected to see collisions occur on a half duplex ethernet network.
The days of using a hub with a half duplex ethernet network is a lot like talking on a conference call where everybody can hear everyone else. You have to only allow one person to communicate on the conference call at a time, or everything becomes jumbled and nobody can make out any information. The CSMA of CSMACD was this process of stations listening on the network to determine if they were able to communicate. The CD part of the CSMACD was the stations able to recognize that a collision occurred, cleared out the network with the jammed signal, and try to retransmit again.
We used to connect ethernet networks together with coax. Everyone was literally on the same cable. So, of course, every station heard all of the other stations communicating on the network. With ethernet hubs, this changed the way the network looked, because we were sending all of our wires down to one central device. But the devices still all heard each other communicating, very similar to the way it worked on a coax network.
With this hub in the middle, two stations could still communicate at the same time and a collision would occur. The only way to segment out these different networks, or segment out these collision domains, was to put a switch or bridge in the middle of the communication. With a switch in the middle, everyone on one side of the network would be one collision domain, and everyone on the other side of the network would be a different collision domain.
This meant, as our networks grew larger and larger, we could continue to segment them to limit the impact the collisions would have on a particular group of devices. And on today’s modern networks, we’ve effectively removed collision domains by adding a switch into the middle of the network and enabling full duplex communication between all of these devices. That means that everyone is able to both send and receive simultaneously, without any concern that a collision might occur.
Broadcast frames are completely different than dealing with collisions. Broadcast frames are things that are a necessary evil on a lot of our networks. For example, there are some things that a device may want to send out to all of the other devices on the network. And it does that by sending this as a broadcast address. This might be ARP requests, the operating system may be notifying other devices on the network of operating system functions, and some dynamic routing protocols will use broadcast domains to advertise the routes that are available on the network.
With broadcast domains, everyone who is on a switch network will be able to see that broadcast as it’s sent across the network. A bridge or a switch will simply pass these broadcasts to every other interface on that switch. The only way to block a broadcast is to put a router on the network, and the broadcast will not go through a routed connection.
This is a good reason to keep networks a convenient size. The larger these networks become, the more broadcasts will appear on the network. If you segment the network, then you’re probably going to be segmenting the amount of broadcast that would occur on a particular broadcast domain. Here’s our modern, full duplex switch network. And every device that is connected to that switch network is on the same broadcast domain.
If this device sends a broadcast frame, it will be redirected to all the other devices that happen to be on this switch network. The way that you would limit the broadcast is to then separate the network with a router in the middle. That means that any devices sending a broadcast on this side of the router will only be seen by these devices. And any device sending a broadcast on this side of the router will only be seen by these devices.