What are the collision domain reducing devices? if collision reduce how it advantage to network?

  • A Switch reduces a collision domains to one domain per port. – user36472 Jul 31 '18 at 6:50
  • 2
    Yeah routers can also reduce collision domains. – user36472 Jul 31 '18 at 10:34
  • 4
    It really depends on what you mean by "collision domain reducing". Looking at it one way, switches reduce the size of collision domains because they increase the number of collision domains, but looking at it the other way, hubs reduce the number of collision domains because they increase the size of collision domains. – Ron Maupin Jul 31 '18 at 13:54
  • 2
    wouldn't it be more accurate to say switches increase Collision domains, since it makes a collision domain for every port connected? – StarThorn Jul 31 '18 at 18:21
  • 2
    Smells like homework – Ron Trunk Jul 31 '18 at 22:31

A collision is the moment when two or more frames from two or more different devices are in the network at the same time. Collisions are a feature of networks, as ethernet, that were designed to share the same media (collision domain) with a contention-based access method.

When a collision happens, the frames become unintelligible, so the devices that sent the frames have to retry later and other devices in the network have to wait until the collision dissapears to try to use the network.

Having a few collisions is not a problem. However, if 50% of the frames in a network suffer collisions, your network will be slow, and a slow network can produce higher layer problems as application disconnections, errors, etc.

As more devices are connected to a network the chance of having collisions increases, so the best way to reduce collisions is to segment the common media. This segmentation is achieved with switches. Each switch port is a "collision domain", it means that collisions happening in a port are not reproduced in the other ports of the switch.

When your devices are connected to an ethernet hub and device A sends a frame to device B, all the other devices see the frame and are unable to use the network until that frame ends.

When your devices are connected to an ethernet switch and device A sends a frame to device B, the frames go from port A to port B and the other ports don't see the frame so they see the network free and can use it. It means that multiple conversations can be achieved and that collisions happening in one port are isolated to that port.

  • 1
    I'd say the second sentence of this otherwise good answer seems a little off to me. I would say Token Ring networks share the same media but do not have collisions. So it's more like collisions are a feature of network protocols (such as Ethernet) that allow on-demand access to a shared medium. Token Ring only allows access to the medium when the node that wants to send has the token. It's basically a time slot system. You might add that routers separate collision domains and differ from switches in that routers separate broadcast domains also, unlike switches. – Todd Wilcox Jul 31 '18 at 16:50
  • 1
    The term of art that describes why Ethernet can have collisions is that it uses a contention-based access method, while Token Ring uses a channel access method (AKA "deterministic"). So the access method is the aspect of a protocol that determines whether there can be collisions or not on it. – Todd Wilcox Jul 31 '18 at 16:55
  • @ToddWilcox edited !!! – jcbermu Aug 1 '18 at 7:45
  • I might find some word other than segment, because I would tend to think of that as a feature of hubs. – richardb Aug 2 '18 at 8:21

In networks with a shared medium, a collision domain encompasses the nodes connected to it. In each collision domain, only a single node may be transmitting at any given time.

If two nodes start transmitting at the same time (due to a signal's propagation delay, this isn't noticeable immediately) the data is garbled, so they need to cease transmission and follow the protocol before trying again. Usually, there will be some random delay before that, so the network bandwidth isn't utilized in this time. Bandwidth is wasted for the time before the collision detection (the unsuccessful, partial transmission) plus the jamming time (to make sure each node has detected the collision) plus the necessary delay (to make a repeated collision less likely).

The effect can be reduced by separating larger collision domains into smaller ones or by getting rid of collisions completely. A network device capable of buffering a packet/frame enables separation of the source's collision domain from that of the destination - this can be a switch, a bridge, or a router. This reduction in collision domain size decreases the likelihood of collisions and therefore can potentially increase network throughput.

When both sides of a link are capable of buffering, an appropriate link can even be used for transmitting in both directions simultaneously. This is called full duplex and it removes the collision domain completely. This is the common mode of fully switched Ethernet operation today. Wi-fi on the other hand uses a inherently half-duplex medium and cannot use full duplex (at least not currently).

  • So we could define the collision domain as the area of ​​the network where collisions can happen, along with the nodes it affects? – Eduardo Sebastian Apr 22 at 2:40
  • @EduardoSebastian Collions can happen in adjacent collisions domains (independently), but they don't propagate beyond their very own domain. – Zac67 Apr 22 at 7:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.