1

I understand that one of the reasons is that ue to the effects of attenuation the collision detection mechanism is not effective beyond 2500 metres (1.5 miles). Segments cannot sense signals beyond that distance. They might, therefore, not be aware that a computer at the far end of a large network is transmitting. Should more than one computer transmit data onto the network at the same time a data collision will take place that will corrupt the data. What am I missing? Could you please explain the concept with clarity, may be with an example?

Thanks!

Rowana Ravenclaw
  • 13
  • 1
  • 2
  • 4

1 Answers1

4

I think you are confusing two seperate issues here.

For twisted pair and coax Ethernet the length of an individual segment is limited by signal integrity. This limitation has nothing to do with CSMA/CD.

The total size of a collision domain (and the individual segment length for half-duplex fiber systems) is limited by timing issues.

For correct CSMA/CD it is important that a collision is "seen" either everywhere or nowhere. If the receiver sees it as a collision but the sender doesn't then you get lost frames. If the sender sees it as a collision but the receiver receives it successfully you get duplicate frames.

To ensure that everyone sees a collision the minium packet length must be more than twice the propagation delay from one end of the collision domain to the other.

Finally note that with full duplex Ethernet collisions are simply not possible, so CSMA/CD is not needed or used. This allows full duplex fiber Ethernet links to operate over very long distances.

Thanks.. Could you please elaborate on this statement - 'To ensure that everyone sees a collision the minium packet length must be more than twice the propagation delay from one end of the collision domain to the other'?

Consider two hosts, A and B at opposite ends of the network.

A starts to send a packet. The data starts moving through the network towards B.

Just before the first data from A arrives B also starts transmitting. B quickly detects a collision.

Bs transmission starts moving (probablly but not nessacerally in a garbelled in in a garabled form) through the network towards A.

If A is still transmitting when the transmission from B arrives then everyone sees the collision. However if A has stopped transmitting then as far as it is concerned it sent it's data successfully. To avoid this situation the transmission time for the smalled allowed packet must be more than twice the delay from one if the collision domain to the other (including delays inside equipment).

How about explaining what exactly is meant by a length being more than a delay.

Technically I should probablly have said "packet transmission time". Of course for a fixed transmission rate transmission time and packet length (including any header/trailers) have a direct relationship.

plugwash
  • 6,719