More on CSMA/CD



CSMA/CD.  We've already surveyed CSMA/CD in other pages.  Now I want to get into a bit more depth.

I'm reiterating the following list that breaks CSMA/CD into its components.

  • Carrier Sense
    • when a computer wants to transmit it first listens to the cable to see if any other device is currently transmitting
    • it senses the state of the carrier - whether it is busy or not
  • Multiple Access
    • more than one computer can begin transmitting on the network when no other messages are sensed
  • Collision Detection
    • if at least two devices sense that nothing else is being transmitted they can end up sending at the same instant and causing a collision which can be detected by a specified mechanism

When signals collide they are both eliminated from the network.  Then each computer waits a random amount of time and then resends the same signal.  It needs to be a random interval otherwise the devices would likely again transmit simultaneously.

At its basis CSMA/CD is actually a half duplex architecture.  Hopefully, you remember that half duplex implies that only one station in a pair can transmit at any one point in time.  Thus a client station operating based purely on the specifications established for CSMA/CD, without any extra functionality created by other devices, cannot be sending and receiving at the same time.

Some authors compare the CSMA/CD architecture to people communicating properly in a conference call meeting:

  • each person doesn't know when another person is going to speak
  • a participant wanting to say something has to wait for the phone line to become quiet before they can start speaking
  • when the line becomes quiet, it is possible for two or more participants to start speaking at the same time
  • if two people start speaking at the same time, since it is essentially not viable for others to understand, the speakers must stop talking and wait until the line becomes clear again before speaking

In Ethernet specifications the amount of time someone would wait after a collision of trying to speak at the same time is randomly generated.  Obviously, as the number of participants increases, the likelihood of having collisions increases.

The network diameter refers to the distance between Ethernet stations at the extreme ends of a broadcast/collision domain.  The rules for 802.3 Ethernet specify that a collision needs to be detectable within the time it takes to transmit the smallest legal Ethernet frame.  The smallest frame is 64 bytes or 512 bits.  On a 10Mbps Ethernet this implies that the maximum wire length is 2800 meters.  The time it takes for an Ethernet frame to traverse the network diameter is called the Ethernet slot time.

Fast Ethernet.  Fast Ethernet is the name for 100 Mbps Ethernet.  The standard for this faster variant, 802.3u, was created in 1995.  Unfortunately, due to the capabilities for faster transmission speeds, the maximum allowable collision/broadcast domain diameter is 200 meters.  Fortunately, other devices such as switches and routers can be used to aggregate LAN segments.

Full Duplex Ethernet.  Fortunately, IEEE ratified another standard modification in 1995 called 802.3x which allows for full duplex operations.  But full duplex operations can only work in point-to-point environments where there is only one device in a collision domain.  Thus stations connected to hubs, repeaters and the like are unable to operate in full duplex mode.  Stations connected back-to-back or through layer 2 switches are able to operate in full duplex mode.

With full duplex mode operations the bandwidth availability is theoretically doubled.