We are trying to reveal the effect channel errors have on the throughput and in the construction of a collision-free schedule with CSMA/ECAqos.
What we know
We know that by not resetting the backoff stage after a successful transmission we implement Hysteresis in CSMA/ECA. That is, we adapt the length of a collisoin-free schedule according to previous collisions; this in turn provides a very rough estimate of the number of contenders.
Moreover, if we instruct nodes to stick to their current deterministic backoff until suffering from more than one collision CSMA/ECA is able to maintain the collision-free schedule (if any) in the presence of a unideal channel.
In order to overview the effect of an increased stickiness on the construction of the collision-free schedule we execute the simulations under an imperfect channel. This channel has a error probability pc, which affects the channel for intervals of t seconds.
The following figures show the number of collisions seen in the last 1000 slots from the channel perspective (pc = 0.1, t = 10). Each simulation is 100 seconds long, with 4 contenders running CSMA/ECAqos with four Access Categories (AC) in saturation. The Channel errors curve show the number of transmissions that suffered a channel error in the last 1000 slots.
NOTE: that is, if a station is unsuccessful in its transmission attempt it can be because, 1) suffered a collision with other station, therefore the slot will be marked with a collision; 2) a channel error forced the transmitter(s) to assume a collision, nevertheless the slot is marked as containing an error instead of a collision.
What can we see?
As the stickiness increases, the number of collisions originated by a channel error are drastically reduced, as it can be seen in Fig. 3. Nevertheless, an infinite stickiness prevents an effective schedule adaptation, colliding even during the intervals of pc = 0.