Today’s networks are designed with the considerations of broadcast domains and, to some degree, the historical characteristics of collision domains. In this video, you’ll learn about both of those domains and how each of them are associated with today’s network configurations.
<< Previous: Unicast, Multicast, and BroadcastNext: Routing Tables >>
Two networking terms that you may have heard are broadcast domain and collision domain. And although they have domain in their name these really are two very different things. A collision domain is something we used to be very aware of. These days it’s more of a historical footnote in networking. This was a term that we used back in ethernet networks that were not communicating with switches. In those scenarios, we were in a half duplex communication where one person could communicate or they could receive, but you could not do both at the same time.
With today’s full duplex communications and connections to our switches we don’t tend to see collisions any longer. The word collision was also a bit of a bad word to use to describe what was happening on an ethernet network when two devices communicated at the same time. Yes, the communication didn’t work properly, because the collision of information occurred on that single bus. But, it isn’t necessarily a bad thing.
This was a normal part of ethernet communication you expected to see collisions occur, which are little bit different than the negative connotation of the word collision. The reason these collisions occurred is that the entire network was one big ethernet segment and everybody on that segment was communicating at half duplex. That meant that while you were sending information, you could not receive information. Why you were receiving information, you couldn’t send information. It was like everybody being on one big conference call, only one person can talk at a time. And if more than one person starts talking on the conference call, they may not even realize it. Then you have two people communicating. You don’t hear what either of them said and they’re going to have to repeat that information.
So on these half duplex ethernet networks only one device can communicate at a time. And it determined whether it could communicate using something called CSMA, which stands for Carrier Sense Multiple Access. The devices we’re listening for a carrier and there were multiple devices on the network that could communicate at one time.
The other part of the ethernet communication mechanism was the Collision Detection or the CD. Devices when they were sending information needed some way to tell if that information was being collided with someone else who is trying to talk with at the same time. There was a difference on the wire and voltage when that occurred and both stations then realized both of us are communicating simultaneously. We need to now clear the wire by sending something called a jam signal and then everybody waited for a random amount of time. And then could resend that information and try to get it onto the network before anybody else was communicating.
In the early days of ethernet, we communicated over coax networks and there was a single wire that everybody connected to. And that single wire or single ethernet bus was one big collision domain. Everyone could hear everyone else when they were communicating and if more than one device was to send simultaneously, we would have a collision.
When we upgraded to twisted pair we still had a single collision domain, because we were using hubs. A hub is effectively a multipoint repeater, which means we have all these devices plugged in. And when traffic comes into one interface, it’s automatically sent to every other interface on the hub. It was as if we collapse that bus all into one single device and it all now exists inside of this hub. There was no separation of the signal. When one person communicated everybody else on this collision domain could hear that communication.
One way to separate out a collision domain would be to separate the network by using a bridge. These days a bridge would take the form of a switch. So if we added a switch into the middle of the network one side of the network would then be one collision domain and the other side would be the other collision domain. By adding the switch into the middle of the network, we were minimizing the number of devices on a single network segment. Therefore, minimizing the number of collisions that would be occurring on that network segment.
These days every single device is connecting directly to a switch and we’re generally connecting up at full duplex. So there’s no possibility that more than one station could ever be involved with a collision. So because of that, you can think that every device would be on its own collision domain. And even though we call this a collision domain because we’re running at full duplex, we’ve effectively removed the chance of ever having a collision.
A broadcast domain is the size of the network that’s impacted when a broadcast is sent. This is something that is a necessary evil. We need these broadcast to be able to perform our probes over the network or for operating systems to advertise a service. But we have to be concerned about this, because every device on the network is listening for broadcast. It receives the broadcast that are sent and it has to at least examine what’s inside of those broadcast, whether it pertains to that station or not. Unlike a collision domain, a broadcast domain passes right through a switch. If you send one broadcast into a switch, the switch will send that broadcast to every other interface on that switch. A broadcast domain only stops at a router. That is the device that will not allow a broadcast to pass through it.
So if we’re going to examine where our broadcast domains might be, you’ll notice they stop every time there is a router. These are important things to consider, because the more devices you put on a subnet, the more devices you connect to a switch, the more broadcast you’re going to have. And eventually, you’ll find these broadcast create some overhead for every single device on the network. So you want to try to minimize the impact of broadcasts.
If we place a router in the middle of the network, we can separate out these broadcast domains. So on the left side, we have a switch and a network. If they broadcast is set from this network, it stops at the router. So this is one broadcast domain. If broadcasts are set on this network, they would communicate to all the devices on this network. But again, would be stopped by this router. Which means we have two separate broadcast domains, both of them separated by this router right in the middle.