How does the ethernet deal with collisions?

CSMA / CD and collisions (Ethernet)

CSMA / CD (Carrier Sense Multiple Access with Collision Detection) is an Ethernet access method in which several network participants can access the transmission medium. The handling of collisions during signal transmission plays a major role here.

CSMA / CD - Carrier Sense Multiple Access with Collision Detection

  • Carrier Sense (carrier status detection): Each station checks whether the transmission medium is free.
  • Multiple Access: Several stations share the transmission medium.
  • Collision Detection: If several stations transmit at the same time, they will detect the collision.

CSMA / CD process

The original Ethernet corresponds to a bus topology to which several stations are connected (multiple access). It is specified that all stations are allowed to read the signals on the bus, but not send them at the same time. Which of the connected stations is allowed to send is determined by the CSMA / CD procedure, which works according to the "listen before talk" principle.

All stations permanently monitor the transmission medium (carrier sense). You can differentiate between a free and an occupied transmission medium. If the transmission medium is free, it is allowed to send. If a station wants to transmit, it checks whether the bus is free. If it is free, the station begins to transmit. This means that it puts the data signal on the bus.

During the signal transmission, the station checks whether the signal sent is identical to the signal on the bus. If the transmitted signal does not correspond to the intercepted signal, then another station has transmitted at the same time. The two signals are superimposed. This condition on the transmission medium is known as a collision. This collision can be detected by permanently checking the status of the line (collision detection).

If a collision is detected, the transmission is aborted. The transmitter that detects the interfering signal first sends a special signal so that all other stations know that the network is blocked. After a random waiting time it is checked again whether the bus is free. If this is the case, it is sent again. The process is repeated until the data has been transferred without a collision.

If the transfer could be ended without collision detection, the transfer is considered successful. If the data did not reach the recipient for any reason, then this data must be transmitted through protocols such as B. TCP, be requested again. If this occurs more frequently, more data packets are sent. This depresses the effective transmission speed of the network.

Collisions

Basically, collisions are not to be regarded as malfunctions. In listen-before-talk operation, collisions are part of normal operations. However, collisions become a problem when they get out of hand. The number of collisions increases the more stations want to have access to the transmission medium.

Long lines, a large number of stations and repeaters (signal conditioners and amplifiers) result in different signal transit times depending on the location of the feed. They cause a station to determine a free transmission medium and to send its signal, although the signal of another station is already on the way. A collision occurs, i.e. the superposition of two signals.

CSMA / CD results in limit values ​​for the maximum signal propagation time and the frame size (data packet or frame). Both must not be exceeded. As long as the bandwidth of Ethernet is not exhausted by more than 30%, collisions are hardly noticeable. However, as the load on the network increases, so do the collisions. The only thing that helps here is to reduce the number of stations or to divide the entire network into subnetworks.

Because of the collisions, it is not possible to fully utilize the theoretical transmission capacity. In practice, the nominal output is around 70% in the best case. Under less favorable conditions it is less than 30%. Well below that, the network practically collapses. The cause is simple: the more computers are active in a network, the more collisions occur. And so the achievable data throughput is constantly falling.

Collision domain


The collision domain comprises a network or just a sub-network of lines, stations and coupling elements of layer 1 (OSI reference model). In the collision domain, the collisions must reach each station within a certain time. This is the condition for the CSMA / CD method to work. If the collision domain is too large, there is a risk that sending stations will not be able to detect the collision. For this reason, the maximum number of stations in a collision domain is limited to 1,023 stations. In addition, there is a maximum of two pairs of repeaters between any two stations.


In order to be able to adhere to the conditions for a properly functioning collision domain, the 5-4-3 repeater rule was defined: No more than five (5) cable segments may be connected. A maximum of four (4) repeaters are used for this. End stations may only be connected to three (3) segments.

Note: The repeater rule applies to 10Base2 and 10Base5 (coaxial cable network). In twisted pair networks you only have to observe the repeater rule when using hubs. The consistent use of switches and routers avoids the problems caused by the CSMA / CD process.

How can you prevent collisions?

Basically: The fewer stations there are in a collision domain, the fewer collisions can occur.


In order to avoid collisions and to achieve a higher data throughput, a network on layer 2 is divided into several subnetworks (subnetworks). When putting together the subnetworks, it makes sense to combine the stations that communicate a lot with one another in a subnetwork.
If you divide a logical network with switches or bridges, several collision domains arise. Within a collision domain there is then a single station, another switch or a router to another network. Establishing collision domains reduces data loss due to collisions. This in turn reduces the network traffic caused by repetitive transmissions.


If you generally only work with switches, the term collision domain is no longer used. In a switch, each port and the station connected by a cable form its own collision domain. It is a point-to-point connection. The switch ensures that only the data packets are forwarded to the port via which the destination MAC address of the packet can be reached.

Half duplex and full duplex

Half-duplex Ethernet is based on the CSMA / CD method. This is the original Ethernet up to 10 Mbit / s. Full duplex Ethernet is a further development that is referred to as Fast Ethernet and does not use CSMA / CD. All other Ethernet developments also work in full duplex mode. The stations communicate directly with each other via point-to-point connections.
Because Fast Ethernet usually works in full duplex mode and thus dispenses with CSMA / CD, additional flow control is required. There is a separate standard for this: IEEE 802.3x (Flow Control).

Overview: Ethernet

Other related topics:

Product recommendations

Everything you need to know about networks.

Network technology primer

The network technology primer is a book about the basics of network technology, transmission technology, TCP / IP, services, applications and network security.

I want that!

Everything you need to know about networks.

Network technology primer

The network technology primer is a book about the basics of network technology, transmission technology, TCP / IP, services, applications and network security.

I want that!