Channel capture effect

The channel capture effect is a phenomenon where one user of a shared medium "captures" the medium for a significant time. During this period (usually 16 frames)^{[clarification needed]}, other users are denied use of the medium. This effect was first seen in networks using CSMA/CD on Ethernet. Because of this effect, the most data-intense connection dominates the multiple-access wireless channel.^[1] This happens in Ethernet links because of the way nodes "back off" from the link and attempt to re-access it. In the Ethernet protocol, when a communication collision happens (when two users of the medium try to send at the same time), each user waits for a random period of time before re-accessing the link. However, a user will wait ("back off") for a random amount of time proportional to the number of times it has successively tried to access the link. The channel capture effect happens when one user continues to "win" the link.

For example, user A and user B both try to access a quiet link at the same time. Since they detect a collision, user A waits for a random time between 0 and 1 time units and so does user B. Let's say user A chooses a lower back-off time. User A then begins to use the link and B allows it to finish sending its frame. If user A still has more to send, then user A and user B will cause another data collision. A will once again choose a random back-off time between 0 and 1, but user B will choose a back-off time between 0 and 3 – because this is his second time colliding in a row. Chances are A will "win" this one again. If this continues, A will most likely win all the collision battles, and after 16 collisions (the number of tries before a user backs down for an extended period of time), user A will have "captured" the channel.

The ability of one node to capture the entire medium is decreased as the number of nodes increases. This is because as the number of nodes increases, there is a higher probability that one of the "other" nodes will have a lower back-off time than the capturing node.

The channel capture effect creates a situation where one station is able to transmit while others are continually backing off, thus leading to a situation of short-term unfairness. Yet, the situation is long-term fair because every station has the opportunity to "capture" the medium once one station is done transmitting. The efficiency of the channel is increased when one node has captured the channel.

A negative side effect of the capture effect would be the idle time created due to stations backing off. Once one station is finished transmitting on the medium, large idle times are present because all other stations were continually backing off. In some instances, back-off can occur for so long that some stations actually discard packets because maximum attempt limits have been reached.

References

1. ^ Kopparty, S; Krishnamurthy, S. V. Faloutsos, M. Tripathi, S. K. (1998). "Split TCP for Mobile Ad Hoc Networks". GLOBECOM -NEW YORK- 1: 138–142. http://www.cs.ucr.edu/~krish/splittcp.pdf. 

• Ramakrishnan, K. K.; Yang, H. (1994). "The Ethernet Capture Effect: Analysis and Solution". CONFERENCE ON LOCAL COMPUTER NETWORKS 19: 228. http://www.research.att.com/~kkrama/papers/capture_camera.pdf.