2

I learnt why CSMA/CD cannot be used in wireless protocols here. Here is what the answer says:

Wireless transceivers can't send and receive on the same channel at the same time, so they can't detect collisions. This is due to the fact that there's an incredible difference between send power (generally around 100mw) and receive sensitivity (commonly around 0.01 to 0.0001mw). The sending would cover up any possible chance of receiving a foreign signal, no chance of "Collision Detection". For this reason Collision Avoidance with Control Messages is necessary.

On most wired networks the (like Ethernet) the voltage is around 1 to 2.5v; both sending and receiving are roughly the same voltage. So if you're sending a 2.5v signal, and someone else collides with a -2.5v signal, the "Detection" parts will see a signal somewhere around 0v and know a collision occurred.

I have following doubts:

  1. Can we use CSMA/CA (Carrier Sense Multiple Access / Collision Avoidance) on wired networks?

  2. If yes, then is their any protocol to use CSMA/CA on wired network?

  3. If no, why?

  4. I was guessing, it must be technically possible to use CSMA/CA on wired network, but most of them use CSMA/CD, because sending signal and then striving to detect any collision is faster than first ensuring collision will not occur and then sending signal. I am correct with this? Or is there any other reason?

PS: I was trying to add following tags: csma, error-control, link-layer, data-link-layer. But I dont have enough reputation points to create new tags. Please add tags as desired.

3

Can we use CSMA/CA (Carrier Sense Multiple Access / Collision Avoidance) on wired networks?

In theory, yes.

If yes, then is there any protocol that uses CSMA/CA on wired network?

Nothing in use today

If no, why?

I was guessing, it must be technically possible to use CSMA/CA on wired network, but most of them use CSMA/CD, because sending signal and then striving to detect any collision is faster than first ensuring collision will not occur and then sending signal. I am correct with this? Or is there any other reason?

Sounds reasonable to me. but to be sure, you'd have to ask the people who invented Ethernet (Bob Metcalfe) that question. Remember that Ethernet and other protocols went through several stages of development and refinement before becoming what it is today.

4
  • just not so related question: Does CSMA/CA ensure 100% that no collision will occur? (After going through its algorithm, I am still guessing if it gives 100% guarantee of no collision, may be I have to ponder upon it more. But can you confirm?) If answer is NO and if it can involve collision, then why it does not have collision detection mechanism? or it does indeed have such mechanism and I am missing something?
    – anir
    Oct 29 '19 at 12:08
  • No, it's not 100%. Station A and station B could simultaneously detect an idle channel and begin transmitting. This is a very rare occurrence, so we rely on other mechanisms. In a wireless network, you can't distinguish between a collision and any other interference.
    – Ron Trunk
    Oct 29 '19 at 12:12
  • Can you please name which "other mechanisms"? Are they part of CSMA/CA or you are pointing at mechanisms of higher layer protocols (like re-transmissions in TCP)?
    – anir
    Oct 29 '19 at 14:20
  • 1
    Higher layer protocols
    – Ron Trunk
    Oct 29 '19 at 14:53
4

Can we use CSMA/CA (Carrier Sense Multiple Access / Collision Avoidance) on wired networks?

We could in theory, but CSMA/CD efficiency and performance are better. So, for wired networks - where reception is guaranteed for all nodes (no hidden station) and listening while transmitting is possible - there's no reason to use CSMA/CA.

If yes, then is their any protocol to use CSMA/CA on wired network?

No.

I was guessing, it must be technically possible to use CSMA/CA on wired network, but most of them use CSMA/CD, because sending signal and then striving to detect any collision is faster than first ensuring collision will not occur and then sending signal. I am correct with this? Or is there any other reason?

If you look at the CSMA/CD algorithm, there's simply no more efficient way to handle media access for a shared wire when there's no central scheduling instance or a token-like concept.

Note that CSMA/CD and half-duplex Ethernet is obsolete and all but extinct due to the ubiquity of cheap switches. Switched Ethernet uses full-duplex links where there's no way a collision can happen. Also, half-duplex mode isn't supported for anything using Gigabit Ethernet or faster.

13
  • "half-duplex mode isn't supported for anything using Gigabit Ethernet or faster." In practice, 1000Base-T doesn't have half duplex (no hubs exist), but it is written into the standard (after a contentious debate). Half duplex was eliminated from 10GBase-T and above standards.
    – Ron Maupin
    Oct 29 '19 at 12:44
  • @RonMaupin Absolutely - there's simply no hardware for GbE HDX is what I wrote. ;-)
    – Zac67
    Oct 29 '19 at 12:47
  • Well, I dont have significant first hand experience in actually implementing networks in organizations. I have mostly academic networking background. So am stuck with knowledge given by reference books and hence struggling with your your last paragraph.Do you mean to say CSMA/CD is half duplex by nature (after reading your last paragraph, I gave some thoughts to it and now I feel yes. Need confirmation.) How switched Ethernet links are full duplex? Any details / links?
    – anir
    Oct 29 '19 at 14:32
  • 1
    @anir Half-duplex mode requires a method to arbitrate the medium - like CSMA/CD, sender tokens or similar - or CSMA/CA for wireless. Full duplex means each node can send and receive simultaneously, requiring not only a full-duplex capable medium (unlike 10BASE2 or a single wireless channel) but also buffering link aggregators/concentrators - exactly what switches do.
    – Zac67
    Oct 29 '19 at 14:48
  • Just want to confirm my interpretation of your comment plus some more things: Both wireless and wired mediums are half duplex. But CAT cables employ multiple set of twisted pair cables, some used for transmitting and some for receiving, hence CAT cable as a whole acts a full duplex medium. Does that mean those full duplex cables does not require any kind of media access control mechanism (like CSMA)?
    – anir
    Nov 2 '19 at 17:24
0

I don´t think it is possible. CSMA/CA is used where the medium is common to all the devices in tx and rx (like Wifi), that´s why it is half duplex and need to avoid collisions. In the beginning there was a ring network using a coaxial cable and "T" connectors, so collision avoidance and half duplex made sense. Nowadays ethernet cables have separated Tx and Rx lines and computers are connected to a cheap and fast full duplex switch that deals with the connections.

1
  • "In the beginning there was a ring network using a coaxial cable and "T" connectors" No, it was a bus, not connected in a ring, and it had terminators on each end to prevent signal reflection. Connecting in a ring would guarantee a collision on every frame.
    – Ron Maupin
    Jan 22 at 22:12

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.