Effect of stickiness in the throughput

Stickiness helps maintain a collision-free schedule once it is reached. Higher levels of stickiness prolong the collision-free state on the network because increases the resilience of the nodes when faced with channel errors. Nevertheless, when this number is too high (1000 for instance), it requires 1000 consecutive collisions so the node could adapt its schedule (via Hysteresis) therefore prolonging the convergence of CSMA/ECAqos and degrading the throughput due to collisions.

Here we test CSMA/ECAqos with 4 nodes and 4 AC each in saturation (the maximum number of nodes in collision-free CSMA/ECAqos when using EDCA parameters). The channel is modelled as in previous posts, that is, every interval of t seconds the channel has either a probability of error, %pc, or acts as a perfect channel. We then try different levels of %pc and stickiness while keeping t = 20.

TL;DR: lower levels of stickiness allow a fast adaptation of the schedule length. Extremely high levels of stickiness prevent the adaptation of the schedule length, leading to more collisions and lower throughput. Stickiness of 5 allows both the adaptation of the schedule and resilience to channel errors, producing higher throughput for higher levels of %pc than the others levels of stickiness tested.

Stickiness = 1 (a single collision triggers a random backoff)

This will be referred to as the standard level of stickiness in CSMA/ECA.

Fig 1: Avg. Aggregated throughput (stickiness = 1)

Fig 1: Avg. Aggregated throughput (stickiness = 1)

Fig 1: Total collisions and %col of each AC (stickiness = 1)

Fig 2: Total collisions and %col of each AC (stickiness = 1)

 

Stickiness = 5 (five consecutive collisions trigger a random backoff)

Fig 3: Avg. Aggregated throughput (stickiness = 5)

Fig 3: Avg. Aggregated throughput (stickiness = 5)

Fig 4: Total collisions and %col of each AC (stickiness = 5)

Fig 4: Total collisions and %col of each AC (stickiness = 5)

 

Stickiness = 1000 (a thousand consecutive collisions trigger a random backoff)

Fig 5: Avg. Aggregated throughput (stickiness = 1000)

Fig 5: Avg. Aggregated throughput (stickiness = 1000)

Fig 6: Total collisions and %col of each AC (stickiness = 1000)

Fig 6: Total collisions and %col of each AC (stickiness = 1000)

Posted under: CSMA/ECA, Fair Share, Hysteresis, MAC, QoS, WiFi

Leave a Reply