The IEEE 802.3 CSMA/CD protocol is a wired network protocol that is similar in operation to that of IEEE 802.11, but using collision detection instead of collision avoidance. The model is taken from Prism case studies. A PLASMA Lab project containing the models and properties used for this experiment can be downloaded here.
Hypothesis testing
We experiment the smart hypothesis testing algorithm on the following CSMA/CD models and properties:
- CSMA 3 4: model with values K=3 and N=4 and property !”collision_max_backoff” U<=#150 “all_delivered”
- CSMA 3 6: model with values K=3 and N=6 and property !”collision_max_backoff” U<=#150 “all_delivered”
- CSMA 4 4: model with values K=4 and N=4 and property !”collision_max_backoff” U<=#300 “all_delivered”
We use the value 0.01 for the parameters alpha, beta and delta and report the results in the table below. Asterisks denote the true probability of the property. We give the results achieved with the minimum simulation budget. As expected, when the probability threshold of the property get closer to the true probability the analysis needs more simulation, and therefore a greater budget.
| CSMA 3 4 | Proba | 0.5 | 0.8 | 0.85 | 0.86 | 0.9 | 0.95 |
| Budget | 200 | 1000 | 100000 | * | 500 | 500 | |
| Time(s) | 0.5 | 3.5 | 737 | * | 2.9 | 2.5 | |
| CSMA 3 6 | Proba | 0.3 | 0.4 | 0.45 | 0.48 | 0.5 | 0.8 |
| Budget | 500 | 1000 | 10000 | * | 5000 | 500 | |
| Time(s) | 1.3 | 5.2 | 79 | * | 39 | 2.6 | |
| CSMA 4 4 | Proba | 0.5 | 0.7 | 0.8 | 0.9 | 0.93 | 0.95 |
| Budget | 200 | 500 | 500 | 1000 | * | 500 | |
| Time(s) | 0.2 | 0.3 | 4.0 | 8.6 | * | 3.8 |
