A collision domain is a section of a network where data packets can interfere with each other when two or more devices try to send at the same time. If two devices on the same collision domain transmit simultaneously, their signals overlap and corrupt each other, forcing both to stop and try again. This wastes bandwidth and slows everything down. The concept comes from early Ethernet networks where devices shared a single communication channel, and it remains a foundational idea in networking even though modern equipment has largely solved the problem.
How Collisions Actually Happen
In a collision domain, only one device can transmit at any given moment. Every other device on that segment has to listen and wait its turn. When two devices happen to send packets at the exact same time, neither packet reaches its destination intact. Both devices detect the collision, stop transmitting, and wait a random amount of time before trying again.
“Simultaneously” here doesn’t mean at the literal same instant. Because electrical signals travel at a finite speed, a device at one end of the network might start transmitting without realizing a device at the other end already started a fraction of a second earlier. The signal from the first device simply hasn’t reached the second one yet. This is why collision domains have size limits: the network segment has to be small enough that a transmitting device can detect a collision before it finishes sending its packet. If the segment is too large or the packet too small, the sender might finish transmitting and never realize there was a collision at all.
The Protocol That Manages Collisions
Ethernet networks handle collisions using a protocol called CSMA/CD, which stands for Carrier Sense Multiple Access with Collision Detection. The name describes the three things each device does. First, it listens to the channel (carrier sense). If the channel is quiet, it starts transmitting. If another device is already sending, it waits. Second, multiple devices share the same channel (multiple access). Third, if a device detects that its transmission is colliding with another, it immediately stops sending.
After a collision, both devices wait a random interval before retransmitting. The randomness is important because if both waited the same amount of time, they’d collide again. With each successive collision, the range of possible wait times doubles, spreading devices further apart in time and making repeated collisions less likely.
Hubs Create Collisions, Switches Eliminate Them
The type of hardware connecting your devices determines whether collisions are a problem. A hub is the simplest networking device: when one device sends data through a hub, the hub copies that signal to every other port. This means every device connected to the hub shares one collision domain. If device A is talking, devices B, C, and D all have to wait. If B and C both try to send at once, their packets collide.
A switch works differently. It learns which device is connected to each port and sends data only to the intended recipient. When device A sends a message to device B through a switch, devices C and D aren’t affected at all. C and D can carry on their own conversation at the same time with no collision. Each port on a switch is its own separate collision domain. A 24-port switch creates 24 independent collision domains, giving every connected device a dedicated path.
Routers take this a step further. Each port on a router is also a separate collision domain, plus routers separate broadcast domains (the zones where broadcast traffic spreads), which switches don’t do by default.
Collision Domains vs. Broadcast Domains
These two terms often come up together, and they describe different problems. A collision domain is the zone where packets can physically collide. A broadcast domain is the zone where a broadcast message (a message addressed to every device) gets delivered. They’re segmented by different equipment.
- Collision domains are separated by switches and routers. Every switch port is its own collision domain.
- Broadcast domains are separated by routers or by VLANs configured on switches. A basic switch, without VLANs, puts all its ports in one broadcast domain.
A hub creates a single collision domain and a single broadcast domain for all connected devices. A switch breaks up collision domains (one per port) but keeps one broadcast domain. A router breaks up both. Understanding this distinction matters when designing networks because too many collisions slow individual connections, while too much broadcast traffic congests the entire network.
What Happens When Collision Domains Get Too Large
In large collision domains, performance degrades in two ways. First, more devices sharing the same domain means more competition for the channel. Total bandwidth is divided among all devices, so adding more devices to a hub-based network gives each one a thinner slice. Second, collisions themselves waste time. Every collision forces a retransmission, and retransmissions consume bandwidth that could have carried new data.
A particularly nasty problem is a late collision. This happens when a collision occurs after the sending device has already transmitted enough of its packet that it assumes delivery was successful. The sender never detects the collision, so it never retransmits, and the data is simply lost. In older networks, late collisions happened when collision domains were physically too large for signals to make a round trip in time. In modern networks, late collisions are almost always caused by a duplex mismatch, where one side of a link is set to full duplex and the other to half duplex. The full-duplex side sends without checking for collisions, while the half-duplex side sees constant interference. The half-duplex side registers late collisions, and the full-duplex side sees corrupted frames.
Why Modern Networks Rarely Have Collisions
Full-duplex communication has made collisions essentially a thing of the past in well-configured networks. In full-duplex mode, a device can send and receive data at the same time on separate channels, so there’s no competition for the wire. Modern switches operate in full-duplex mode by default, and since each port is its own collision domain with only one device on it, collisions can’t occur.
Hubs, which were the primary source of collision domains, have been obsolete in professional networking for years. Switches are cheaper and faster, and they give every device dedicated bandwidth instead of forcing everyone to share. Gigabit and faster Ethernet standards don’t even support the half-duplex mode that made CSMA/CD necessary.
That said, the concept still matters. It comes up in networking certifications, in troubleshooting legacy equipment, and in understanding why networks are designed the way they are. If you ever encounter a network running slow with unexplained packet loss, checking for duplex mismatches and accidental hub-like behavior is still a real diagnostic step. The collision domain is a solved problem, but knowing how it was solved helps you understand the architecture of every modern network.