# Why does the Throughput of Non-persistent CSMA increase with attempts made to send packets

Its intuitive for Aloha and p-persistent CSMA that the more we try to send during a packet time the more the collisions we get; however, I have been unable to justify why non-persistent CSMA is different.

The only thing it does that 1-persistent CSMA doesn't is repeat transmission after a random time if the channel is busy. It's obvious that we should get more collisions with more packets per packet time because multiple ones would easily sense the channel as free simultaneously.

• Your graph shows the opposite: the less aggressively the channel is accessed, the higher the throughput.
– Zac67
Commented Dec 8, 2022 at 16:14
• Not sure if I'm not wearing glasses of if the curve for "Nonpersistent CSMA" seems to approach 1 as G tends to infinity. Commented Dec 8, 2022 at 16:35
• Non-persistent CSMA isn't aggressive. If the channel is busy it waits a random time before checking again.
– Zac67
Commented Dec 8, 2022 at 18:43
• I see but how then does having more transmissions per packet time result in even better throughput, contrary to all others in the graph? Commented Dec 9, 2022 at 14:41
• I don't understand more transmissions per packet time - successful transmissions increase throughput, collisions decrease throughput. Also, the less aggressively the channel is accessed, the higher the throughput.
– Zac67
Commented Dec 9, 2022 at 14:55

## 1 Answer

If there are average transmissions per packet time that means there's a greater chance of no transmissions per packet time. To increase the throughput there needs to be exactly one transmission.

The difference between 1-persistent CSMA and non-persistent CSMA is that 1-persistent CSMA will transmit as soon as the sensed packet has finished - which creates a chance that two nodes will wait for the same packet to finish and transmit at the same time. With non-persistent CSMA, nodes always wait for random amounts of time so the risk of two nodes trying to transmit at the exact same time is low. One will almost always transmit first, and then the other one will sense it and wait.

• Thanks but how does more transmissions help increase throughput? This is what doesn't make sense for me. Commented Dec 9, 2022 at 20:32
• @Essam as I understand it, the chance of a collision is very low because each node waits for a random delay. This graph might be based on an idealized calculation, though, because in reality two random delays that are close together could still cause a collision. Commented Dec 9, 2022 at 20:49
• @Essam transmissions transmit data. If there are no transmissions at all then no data is transmitted and the throughput is 0 Commented Dec 9, 2022 at 21:03