Access technique CSMA/CD

Access technique CSMA/CD

Ethernet networks use a media access technique known as carrier-sense-multiply-access with collision detection (CSMA/CD).

This technique is employed solely in networks with a standard bus (which includes the radio networks that gave rise to the current method).

All computers in such a network have direct access to a standard bus, thus often want to transfer information between any 2 network nodes.

The simplicity of the schematic is one in every of the factors behind the success of the local area network customary.

the cable in stations area unit connected operates in multi-access mode (multiply-access, MA).

With the described approach, a situation is possible when two stations simultaneously try to transmit a data frame over a common cable (Fig. 1).

To reduce the likelihood of this situation, immediately before sending the frame, the transmitting station listens to the cable (that is, receives and analyzes the electrical signals occurring on it) to detect if a data frame from another station is already transmitted over the cable.

If the carrier is recognized (carrier-sense, CS), then the station postpones the transmission of its frame until the end of someone else's transmission, and only then tries to transmit it again.

But even with this algorithm, two stations can simultaneously decide that there is no transmission on the bus at a given time, and start simultaneously transmitting their frames.

A collision is said to occur as the contents of both frames collide on a common cable, resulting in a distortion of the information.

In order to correctly handle a collision, all stations simultaneously monitor the signals that appear on the cable.

If the transmitted and observed signals differ, then a collision detection (CD) is detected.

To increase the probability of immediate detection of a collision by all stations in the network, the collision situation is enhanced by sending to the network by stations that have begun transmitting their frames a special bit sequence called a jam sequence.
Upon detecting a collision, the transmitting station is required to stop transmitting and wait for a short random time interval, and then may attempt to transmit the frame again.

From the description of the access method, it can be seen that it is of a probabilistic nature, and the probability of successfully obtaining a common environment at its disposal depends on the network load, that is, on the intensity of the need for frame transmission in stations.

When developing this method, it was assumed that the data transfer rate of 10 Mb / s is very high compared to the needs of computers in mutual data exchange, so the network load will always be small.

This assumption often remains true to this day, but real-time multimedia applications have already emerged that require much higher data rates.

Therefore, along with classic Ethernet, there is a growing need for new high-speed technologies.

The CSMA/CD method defines the basic timing and logical relationships that guarantee the correct operation of all stations in the network:

• A pause of 9.6 µs must be maintained between two sequentially transmitted information frames on a common bus; this pause is needed to reset the network adapters of the nodes, as well as to prevent the exclusive capture of the data transmission medium by one station.

• When a collision is detected (the conditions for its detection depend on the physical medium used), the station issues a special 32-bit sequence (jam-sequence) into the environment, which enhances the collision phenomenon for more reliable recognition by all network nodes.

• After a collision is detected, each node that transmitted a frame and encounters a collision tries to retransmit its frame after some delay.

• The node makes a maximum of 16 attempts to transmit this frame of information, after which it refuses to transmit it.

• The delay value is chosen as a uniformly distributed random number from an interval whose length increases exponentially with each attempt.

• Such an algorithm for choosing the delay value reduces the probability of collisions and reduces the intensity of issuing frames to the network when it is highly loaded.

Figure1: collision incidence within the CSMA/CD random access technique

(tp - signal propagation delay between stations A and B)

Accurate recognition of collisions by all stations of the network could be a requirement for the proper operation of the LAN network.

If any transmission station doesn't acknowledge the collision and decides that the information frame was transmitted properly by it, then this information frame is lost, since the frame data are distorted because of signal overlap throughout the collision, it'll be rejected by the receiving station (most seemingly because of confirmation mismatch).

Of course, possibly the disconnected data are retransmitted by some upper-layer protocol, for instance, transport or application, operating with association institution and list of their messages.

But the retransmission of the message by the higher layer protocols can occur once a far longer quantity (tens of seconds) compared to the time unit intervals that the LAN protocol operates on.

Therefore, if collisions aren't dependably recognized by the nodes of the LAN network, this can result in a comprehensible decrease within the helpful turnout of this network.

All parameters of the Ethernet protocol are selected in such a way that during normal operation of the network nodes, collisions are always clearly recognized.

It is for this that the minimum length of the frame data field must be at least 46 bytes (which, together with the service fields, gives the minimum frame length of 72 bytes or 576 bits).

The length of the cable system is chosen in such a way that during the transmission of a frame of the minimum length, the collision signal would have time to propagate to the farthest network node.

Therefore, for a data transfer rate of 10 Mb / s, used in Ethernet standards, the maximum distance between any two network nodes should not exceed 2500 meters.

As the frame rate increases, as occurs in new standards based on the same CSMA/CD access method, such as Fast Ethernet, the maximum network length decreases in proportion to the increase in transmission rate.

In the Fast Ethernet standard, it is 210 m, and in Gigabit Ethernet it is limited to 25 meters.

Regardless of the implementation of the physical medium, all Ethernet networks must satisfy two restrictions related to the access method:

the maximum distance between any two nodes should not exceed 2500 m,

the network should not have more than 1024 nodes.

In addition, each version of the physical environment adds its own restrictions to these restrictions, which must also be met.

Let us clarify the main parameters of the operations of transmitting and receiving Ethernet frames, briefly described above.

A station that wants to transmit a frame must first pack the data into a frame of the appropriate format using the MAC node.

Then, to prevent mixing of signals with signals from another transmitting station, the MAC node should listen for electrical signals on the cable and, if a carrier frequency of 10 MHz is detected, delay the transmission of its frame.
After the end of the transmission over the cable, the station must wait for a small additional pause, called the interframe interval (interframe gap), which allows the destination node to receive and process the transmitted frame, and then start transmitting its frame.

Simultaneously with the transmission of frame bits, the node's transceiver monitors the bits received over the common cable in order to detect a collision in time.

If no collision is detected, then the entire frame is transmitted, after which the node's MAC layer is ready to accept the frame from the network or from the LLC layer.

If a collision is detected, then the MAC node stops transmitting the frame and sends a jam sequence that reinforces the collision state.

After sending a jam sequence to the network, the MAC node pauses randomly and re-attempts to transmit its frame.

In the case of repeated collisions, there is a maximum possible number of attempts to retransmit a frame (attempt limit), which is 16.

When this limit is reached, a frame transmission error is detected, the message about which is passed to the upper layer protocol.

In order to reduce the intensity of collisions, each MAC node randomly increases the duration of the pause between attempts with each new attempt.

The time schedule for the duration of the pause is determined based on the truncated binary exponential backoff algorithm.

The pause is always an integer number of so-called delay intervals.

The backoff interval (slot time) is the time during which the station is guaranteed to know that there is no collision in the network. This time is closely related to another important network timing parameter, the collision window.

The collision window is equal to the time it takes for a signal to travel twice between the most distant nodes in the network, the worst case delay in which a station can still detect that a collision has occurred.

The backoff interval is chosen to be equal to the value of the collision window plus some additional delay value to ensure:

• backoff interval = collision window + additional delay
  In 802.3 standards, most time intervals are measured in the number of inter-bit intervals, the value of which for a bit rate of 10 Mbps is 0.1 µs and is equal to the transmission time of one bit.

• The backoff interval value in the 802.3 standard is defined as 512 bit intervals, and this value is calculated for a maximum coaxial cable length of 2.5 km.

• The value 512 also determines the minimum frame length of 64 bytes, since with frames of a shorter length, the station may transmit the frame and not have time to notice the occurrence of a collision due to the fact that the signals distorted by the collision will reach the station in the worst case after the transmission is completed.
  Such a frame will simply be lost.

• The pause time after the Nth collision is assumed to be equal to L delay intervals, where L is a random integer uniformly distributed in the range [0, 2N].

• The value of the range grows only up to the 10th attempt (recall that there can be no more than 16 attempts), and then the range remains equal to [0, 210], that is, [0, 1024].

The values of the main parameters of the 802.3 frame transmission procedure are given in Table 1.

Given the above parameters, it is not difficult to calculate the maximum performance of an Ethernet segment in units such as the number of transmitted packets of minimum length per second (packets-per-second, pps).

The number of Ethernet packets processed per second is often used to indicate the internal performance of bridges and routers, which introduce additional delays in the exchange between nodes.

Therefore, it is interesting to know the net maximum performance of an Ethernet segment in the ideal case, when there are no collisions on the cable and no additional delays introduced by bridges and routers.

Since the packet size of the minimum length together with the preamble is 64+8 = 72 bytes or 576 bits, it takes 57.6 µs to transmit it. Adding an interframe interval of 9.6 µs, we obtain that the minimum burst period is 67.2 µs.

This corresponds to the maximum possible Ethernet segment bandwidth of 14880 p/s.

Thank you