Table of Contents
Understanding Collision Domains
In Ethernet networking, especially in older or shared-media environments, a collision domain is a fundamental idea that explains when and where devices can interfere with each other’s transmissions.
A collision occurs when two devices try to send Ethernet frames on the same shared medium at the same time. The area of the network where such a collision can happen is called a collision domain. Every device inside that area must compete for the right to send data, and when two send at once, both transmissions are damaged and must be resent.
How Collisions Happen
Traditional Ethernet with hubs or a single shared cable operates like many people talking on the same walkie talkie channel. Only one person can talk clearly at a time. If two speak simultaneously, nobody hears a clear message. Both must stop, wait, and try again.
On such a shared Ethernet segment, all attached devices listen to the same electrical or logical signal. Because there is a single shared channel, there is a chance that two devices decide to transmit at nearly the same instant, which creates a collision.
In half duplex Ethernet, the network interface card (NIC) both sends and listens on the same medium but cannot send and receive at the same moment without risking a collision. This context is where collision domains are relevant.
Collision Domains and Ethernet Devices
Different types of Ethernet devices affect collision domains in different ways. The key idea is whether a device repeats the same signal to everyone or separates traffic into distinct point to point links.
The following table summarizes how common devices interact with collision domains:
| Device type | OSI layer (main) | Effect on collision domains |
|---|---|---|
| Repeater | 1 (Physical) | Does not break collision domains. All ports are one collision domain. |
| Hub | 1 (Physical) | All ports share one big collision domain. |
| Bridge | 2 (Data Link) | Breaks collision domains. Each port is a separate collision domain. |
| Switch | 2 (Data Link) | Each port is its own collision domain. |
| Router | 3 (Network) | Also separates collision domains on each interface, and separates broadcast domains. |
A hub simply repeats any incoming signal to all other ports. Because the same electrical signal is shared, every device connected to that hub is part of the same collision domain.
A switch, in contrast, uses MAC addresses and creates a dedicated logical link between two ports for each conversation. This separation isolates collisions. In practice, each switch port is its own collision domain.
Routers connect different IP networks and also sit at boundaries where both collision domains and broadcast domains are separated. The broadcast domain aspect belongs to other chapters, so here you only need to remember that each router interface is at least a separate collision domain.
Collision Domains with Hubs and Shared Media
With a hub based network, all connected devices share the same collision domain. If one device transmits, all others must remain silent until that transmission is complete. Any attempt by a second device to send at the same time will cause a collision that affects everyone.
In this kind of network, Ethernet uses a method called carrier sense multiple access with collision detection, or CSMA/CD, to reduce and handle collisions. The details of CSMA/CD belong elsewhere, but for collision domains the important point is that all participants must coordinate their access to the same shared medium.
As the number of devices in a single collision domain increases, the probability of collisions increases. More collisions mean more retransmissions and less effective throughput. At some point, collisions can dominate, and the network becomes slow and inefficient even if the raw bandwidth of the cable is high.
Collision Domains with Switches
Switches largely solve the problem created by large collision domains. Each switch port behaves as a separate collision domain. When a device is connected to a dedicated switch port, and both sides use full duplex, collisions do not occur on that link.
If you have eight devices each plugged into its own switch port, you now have eight separate collision domains instead of one. Traffic between device A and B does not expose devices C and D to the risk of collision, because those pairs use different switch ports.
This separation is one of the key reasons that switches replaced hubs in modern Ethernet networks. It allows more simultaneous communication and better use of the available bandwidth.
Full Duplex and the End of Collisions
Full duplex means a device can send and receive at the same time over separate channels, one for each direction. In a full duplex, point to point link between a NIC and a switch port, there is no shared channel that multiple devices compete for. There is only one NIC and one switch interface on that link.
Because there is only a single sender on each direction of the link, collisions cannot occur. In other words, collisions are a concept that only applies when multiple devices share the same path for sending traffic. In a full duplex switched Ethernet network, each link has exactly one sender in each direction, so the collision domain for that link effectively contains only one potential transmitter at a time.
This is why modern switched networks using full duplex often treat collisions as a historical concern. However, understanding collision domains is still important because it explains why certain designs scale badly and why older half duplex or hub based segments behave differently.
Measuring and Counting Collision Domains
You often need to be able to look at a small network diagram and identify how many collision domains it contains. The basic rules are simple.
Each hub or repeater does not reduce the number of collision domains. All devices connected through hubs and repeaters that share the same continuous physical segment belong to a single collision domain.
Each switch port is a separate collision domain. Each router interface is also a separate collision domain.
Here are some simple examples described in words.
If three PCs are connected to a single hub, which in turn connects to a router, there are two collision domains. All three PCs plus the hub form one collision domain. The link between the hub and the router forms a second collision domain.
If three PCs each connect to their own switch port, and the switch connects to a router, there are four collision domains. Each PC to switch link is one collision domain, which makes three. The switch to router link is the fourth.
If two switches are connected together, every link between a device and a switch, or between switches, is its own collision domain. The effect scales linearly with the number of active ports.
Collision Domains vs Broadcast Domains
Collision domains and broadcast domains are often confused because both describe where certain traffic types can propagate. The two concepts are different.
A collision domain is about where simultaneous transmissions can interfere with each other. A broadcast domain is about where broadcast frames or packets, for example Ethernet broadcasts, can be received.
A switch breaks up collision domains at each port, but does not break broadcast domains by default. Broadcasts on one switch port are forwarded to all ports in the same VLAN. A router, however, breaks both collision domains and broadcast domains at each interface.
The detailed study of broadcast domains and VLANs belongs to other chapters. Here, keep in mind that one device can change one type of domain but not the other. For example, a switch reduces collision domains significantly, while leaving the broadcast domain unchanged.
Impact of Large Collision Domains
Large collision domains cause performance problems. As more devices share a single collision domain, more of them attempt to send data. Even if each device sends infrequently, the total chance that two happen to send at the same time increases.
When collisions increase, every affected frame must be resent. This repetition wastes bandwidth and causes delays. The real throughput that users experience drops, and the network can feel slow or unreliable.
In very busy shared networks, the effective throughput can become far lower than the theoretical capacity of the cable. For example, a 10 Mbps shared segment might deliver far less useful bandwidth because it is constantly handling collisions and retransmissions.
By breaking a large collision domain into many smaller ones, usually with switches, you reduce the number of devices that compete on any single path. This is why careful control of collision domains is a basic design step when improving network performance.
Designing to Minimize Collisions
Modern best practice is to remove shared collision domains wherever possible by using switches, full duplex links, and avoiding hubs. Each user or server should ideally have a dedicated switch port operating in full duplex. Inter switch and switch to router links should also be full duplex whenever supported.
If a hub or half duplex segment must be used, for example in a legacy environment, it is wise to limit the number of devices in that collision domain and to keep high traffic devices separate through dedicated links when possible.
Switches can also be used to segment an older shared environment into several smaller collision domains. For instance, instead of a single hub with 24 ports, a design might use several small hubs each connected to a switch. Each hub then becomes a separate collision domain instead of combining all devices into one very large domain.
Summary of Key Ideas
A collision domain is the part of a network where a collision can happen if two devices send at the same time. Collisions are inherent to shared, half duplex Ethernet, and are handled through methods such as CSMA/CD, but they reduce performance when they are frequent.
Hubs and repeaters create large shared collision domains. Switches, bridges, and routers divide the network so that each port or interface becomes its own collision domain. In full duplex point to point links, collisions no longer occur, which is why collision domains are mainly a concern in older or special cases.
Although collisions are rare in modern well designed switched networks, understanding collision domains remains essential for reading network diagrams, troubleshooting performance issues on legacy segments, and appreciating how Ethernet has evolved.