Carrier Sense Multiple Access / Collision Avoidance
The English term Carrier Sense Multiple Access / Collision Avoidance ( CSMA / CA ) (multiple access with carrier check and collision avoidance) describes a principle for collision avoidance when several network stations access the same transmission channel . It is often used, among other things, in wireless networks ( wireless LANs ), but is also used in a modified form in communication processes such as ISDN , or in many communication networks in which several clients place data on a bus and collisions must not occur. The problem does not occur in centrally coordinated communication networks.
Motivation for CSMA / CA in radio networks
Wireless networks differ from wired networks in terms of shared media access in two important ways:
- The mains adapter is not necessarily full duplex capable. The medium cannot be monitored during its own transmission . The use of a "collision detection" mechanism, such as that provided by CSMA / CD and used with Ethernet , would then fail. That is why CSMA / CD has been further developed into a mechanism that more consistently follows the principle of “listen before talk” (“first hear, then speak”). Instead of collision detection ("CD - Collision Detection"), the (best possible) collision avoidance ("CA - Collision Avoidance") should take place. Simultaneous data transfers cannot be completely prevented, but can be minimized.
- The range of the signal is very limited, as the reception level decreases with the square of the distance. This can lead to effects such as “hidden” or “exposed” end devices.
Hidden station problem
A Hidden Station or a Hidden Terminal ( engl. "Hidden terminal") in asynchronous and not centrally coordinated referred to communications networks , radio networks or computer networks undesired fact that, during a transmission between two devices (A and B) another potential transmitter (C the hidden terminal) is in the vicinity of the receiver (B), which cannot be seen by the actual transmitter (A) (seeing here means being recognized by Carrier Sense ). This potential sender (C) can disrupt the communication of the other two nodes (A and B) by also sending a message to the node in the middle (B), which can lead to a collision at the receiver (B).
The figure shows the typical situation of the hidden station problem in radio networks with only one modulation method on a radio frequency. Network node A sends data to node B. The signal reaches B, but not the more distant node C due to the limited range of the radio link . C is therefore not aware of the ongoing data exchange and gains the impression that the communication medium is free to send data to node B. If C now also starts sending, the signals at B are superimposed. The result is a data collision due to which B neither successfully receives the messages from A nor from C.
Special procedures cannot exclude this collision, but they create a deterministic behavior of the participants with stochastic procedures. The use of RTS / CTS tries to avoid the hidden station problem. If B replies to a request-to-send from A with a clear-to-send , C overhears this and waits for the time of the transfer between A and B. This cannot completely prevent the problem.
Exposed station problem
Under a Exposed station or Exposed Terminal (in German: exposed terminal ) is, unless in our present scenario, the station sends B to A and C now wants to send to any other station that is not within range of B. C recognizes the signals from B and waits until the transmission between B and A is over. However, since the radio waves from C cannot reach station A at all, there would be no need to wait at all: with A, no conflict could arise at all. Nevertheless, C is dependent on the broadcast of the other two stations (delivered). Waiting unnecessarily wastes bandwidth.
If, however, as is usual with unicast transmissions, acknowledgments (ACKs) are used to confirm correct receipt, the recipient of the message becomes the sender of an ACK after receiving the message. So that the original sender of the message can receive this ACK, stations that are within its range must also wait. In the example, A would confirm the transmission of B with an ACK. If C were to send at the same time, there could be a collision between A's ACK and the message sent by C at B.
If a device wants to send data using the CSMA / CA method, the following sequence is possible:
- First the medium is listened to ("listens", "carrier sense" ).
- If the medium is free for the duration of a DIFS , a backoff time is extracted from the contention window and sent when this has expired.
- If the medium is occupied, the backoff is stopped until the Network Allocation Vector (NAV) has expired , before it continues to run accordingly after another DIFS.
- After the packet has been completely received, the receiver waits for a SIFS before the ACK is sent.
- A collision due to the simultaneous running of the backoff leads to an ACK timeout - after which an EIFS is waited before the entire process can be repeated (DIFS → BO ..).
In addition, procedures are defined that offer a systematic solution for the problem cases of the hidden and exposed station without additionally restricting the transmission level. The prerequisite for this process is that the recipient, who hears both transmitters, intervenes in the process by sending it himself:
- A transmission process is not started while a transmission is running. Each transmitter only broadcasts for a limited time.
- The transmission process is aborted as soon as the transmitter detects a collision by receiving another transmitter. The next transmission is then delayed by a randomly determined pause.
- The receiver, which alone detects the collision, sends a signal in the expectation that both colliding transmitters will recognize this and then both initiate the pause routine.
RTS / CTS coordination
In order to reduce the problem of hidden stations , an extension is defined which is referred to as CSMA / CA RTS / CTS ( R equest T o S end / C lear T o S end). Synonymous with the term MACA for M ultiple A ccess with C Olli sion A voidance.
After waiting for DIFS, the transmitting station tries to occupy the channel with an RTS packet for a certain time specified in the packet. After waiting for SIFS, the recipient confirms this with a CTS packet, which also contains a period of occupancy for the channel.
All stations in the transmission area that receive this RTS remain silent until the CTS response (clear to send, contains the length of the data frame copied from the RTS) returned by the receiver has been received without conflict and the sending station has sent the data. Accordingly, all recipients of the CTS wait according to the length in the CTS.
One advantage of this method is that collisions are only possible while RTS or CTS packets are being sent; the disadvantage is the high cost of exchanging reservation messages. RTS / CTS is particularly suitable for partially solving the hidden station problem. However, two devices hidden from each other can still send RTS to a common receiver at the same time. These collide and are lost. In this case the transmitters wait a random period of time and send RTS 'again. The algorithm used is the binary exponential backoff .
An experimental comparison shows that RTS / CTS packets are worthwhile in every traffic class. The use of the RTS / CTS mechanism is also profitable for short audio frames, which produce a high overhead of RTS / CTS frames.
The problem of the exposed station is not solved by this standard, but it cannot be classified as dramatic as the hidden station problem, since it only leads to a lower throughput.
A further development for ad-hoc networks is the PAMAS protocol (Power Aware Multi-Access Protocol with Signaling Ad Hoc Networks).
The difference to MACA is that receivers that hear the RTS / CTS signal switch themselves off for the duration of the transmission. This is possible because the RTS / CTS each contain the packet size. Furthermore, energy is saved by the fact that the nodes switch off at regular intervals.
Compared to MACA, PAMAS has energy savings of up to 70%. This depends on the degree of networking. The disadvantage of MACA and PAMAS is the lower throughput rate - caused by the restriction of transmissions to only one in a neighborhood of mutually accessible nodes.
For this, the access point sets the next contention period (CP, C ontention P a contention free period (CFP, eriod) C ontention F ree P a eriod). In the CFP, after waiting for PIFS , the AP requests all stations to send a data frame . A station willing to send can start sending after waiting for SIFS , then the AP continues sending the poll packets after waiting for SIFS.
CSMA / CA PCF is optional and is therefore rarely implemented.
- R. Michael Buehrer: Code Division Multiple Access (CDMA). 1st edition. Morgan & Claypool, 2006, ISBN 1-59829-040-1 .
- Volker Jung, Hans-Jürgen Warnecke (Hrsg.): Handbook for telecommunications. Springer Verlag, Berlin / Heidelberg 1998, ISBN 3-642-97703-0 .
- Mark A. Dye, Rick McDonald, Antoon W. Rufi: Networking Basics. ' CCNA exploration companion guide. Addison + Wesley Verlag, Munich 2008, ISBN 978-3-8273-2685-0 .
- Dirk Traeger: LAN practice of local networks . Springer Fachmedien, Berlin / Heidelberg, ISBN 978-3-519-06189-2 .
- Alan Colvin: CSMA with collision avoidance. 1983. doi: 10.1016 / 0140-3664 (83) 90084-1
- DataLink Layer - Carrier Sense Multiple Access with Collision Avoidance (accessed July 27, 2017)
- Hermann Pommer: Roaming between Wireless Local Networks . VDM 2008, ISBN 978-3-8364-8708-5 , p. 179.