2
\$\begingroup\$

Ethernet uses a refinement of ALOHA, known as Carrier Sense Multiple Access (CSMA), which improves performance when there is a higher medium utilisation. When a NIC has data to transmit, the NIC first listens to the cable (using a transceiver) to see if a carrier (signal) is being transmitted by another node. This may be achieved by monitoring whether a current is flowing in the cable (each bit corresponds to 18-20 milliAmps (mA)). The individual bits are sent by encoding them with a 10 (or 100 MHz for Fast Ethernet) clock using Manchester encoding. Data is only sent when no carrier is observed (i.e. no current present) and the physical medium is therefore idle. Any NIC which does not need to transmit, listens to see if other NICs have started to transmit information to it.

Reference 1

I am getting confused about CSMA/CD (half-duplex communication) after reading below sentence.

Data is only sent when no carrier is observed (i.e. no current present) and the physical medium is therefore idle.

Is not pulse (link integrity test pulse, normal pulse, fast pulse) is generated when the medium is idle?

One more thing.

In the output from the show interfaces fastethernet , show interfaces gigabitethernet or show interfaces tengigabitethernet command observe the value of the carrier signal counters. The lost carrier counter shows the number of times that the carrier was lost during transmission. The no carrier counter shows the number of times that the carrier was not present during transmission.

Reference 2

Is carrier lost determined by RX of interface?

\$\endgroup\$
2
\$\begingroup\$

Beginning from 10BASE-T, there is no CDMA/CD in Ethernet because (and while) the underlaying PMA/PMD provides the PCS (providing the service interface for the MAC) with a full duplex data transmission mechanism. In 10BASE-T, 100BASE-TX, 1000BASE-X, 1000BASE-T, etc., the half duplex mode of operation, i.e. providing CRS and CD signals by the PHY to its MAC, is speculatively simulated in the PHY by itself like the following:

  • carrier sense := rx activity or tx activity
  • collision detect := rx activity and tx activity

(Be careful here and remember that rx activity as well as tx activity are on the medium side while carrier sense is on the PHY service interface, i.e. MAC, side.)

Returning to your counters, they (could) show how the actions over the PHY (being taken by the MAC serviced) correlate/correspond with the actions under the PHY (happening at the medium) and may (could) be like the following:

  • lost carrier counter increments when carrier sense transits from HIGH to LOW while TX_EN is HIGH (generally speaking, it's implementation dependent and could be sole TX_EN, sole RX_DV, or combination thereof)

  • no carrier counter increments when TX_EN transits from LOW to HIGH while carrier sense is LOW (it's also implementation dependent and could be other any)

In other words, the first counter indicates how many times the link fails while a packet was transmitted and/or received---collectively, presented---at the medium, and the second counter indicates how many times the MAC attempts to send a packet while there was no physical link.

An yes, in 10BASE-T (and in EEE today) FLPs are generated while the medium is idle, but this is not the case as it is shown above.

\$\endgroup\$

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.