Draft: Better wraparound model integration for ns-3.46
This will eventually replace !2330 (merged).
Here are some preliminary results of the wraparound model (MR !2330 (merged) and !2452),and how this model works with typical 5G-LENA simulation. We compared 3 setups:
- 0th + 1st ring without wraparound,
- 0th + 1st ring with the wraparound model, and
- (as a representative of a traditional way) simulating 3 rings (0th, 1st, 2nd and 3rd), but filtering out results, and only showing the results of the UEs belonging to the 0th and the 1st ring. Notice that the equivalent of the second option would be at least 5 rings, but such simulation in the selected setup with the full buffer traffic, and high number of antennas at gNB would is almost not feasible (8x8x2 (dual polarized)).
Simulating rings without wraparound model looks like this:
And by using the wraparound model, without simulating these additional ring, we get even better "wrapping":
Screenshot from the paper: "R. S. Panwar and K. M. Sivalingam, "Implementation of wraparound mechanism for system level simulation of LTE cellular networks in NS3," 2017 IEEE 18th International Symposium on A World of Wireless, Mobile and Multimedia Networks (WoWMoM), Macau, 2017, pp. 1-9, doi: 10.1109/WoWMoM.2017.7974289."
Results:
In this particular setup, the wraparound model reduces execution time 6 times! Simulation time drops from 8.33 hours to 1,38 hours. If compared with 5 rings, the improvement would be much much higher. In this simulation campaign the results are obtained by repeating each simulation setup 24 times for different RngRun number. Each gNb has only 4 users... while the full outdoor setup should have 10 users. Here are also the throughput CDF curves:
Based on these preliminary results, wraparound models appear essential for network simulators, particularly in large-scale simulations such as these hexagonal scenarios.