QUIC/inet/examples/wireless/throughput/README

This example measures the IEEE 802.11 wireless network throughput from a
number of hosts to the access point (AP). The theoretical maximum
throughput can be calculated using the attached Excel sheet. The host and
the access point transmitter powers are configured so that every host hears
all the others.

1. Single-host throughput

Throughput is measured by the "sink" submodule of the AP. It is recorded
into the output scalar file, but can also be inspected during runtime.

The Excel sheet includes throughput measured by the simulation, and compares
it to the theoretical maximum which is roughly 5.12 Mbps (at 11 Mbps bitrate
and 1000-byte packets). The theoretical value and the simulation output
are very close, the difference being less than 1 kbps.


2. Multi-host throughput

In this case, the average backoff interval on the channel is smaller which
would increase throughput, but are also collisions. This makes the
calculation of the theoretical maximum rather complicated. To simplify
things, the Excel sheet does not consider collisions when calculating the
maximum throughput. For 3 hosts the minimum backoff period is (n - 1) ^ 2 /
(4 * n) where n is the contention window size, that is, n=32. This shows an
average contention window of roughly 7.5, and an increased total channel
throughput.

The difference between the theoretical throughput maximum and the measured
throughput is caused by collisions. Adding the number of frame collisions
to the number of correctly received frames makes the difference less than
50 kbps. The difference is probably caused by inter-frame spacing: the
simulation waits EIFS after detecting a collision, but the spreadsheet
calculates with DIFS.

The experiments are were inspired by the following paper:
S. Choi, K. Park and C. Kim, "On the Performance Characteristics of WLANs:
Revisited", Proceedings of the ACM SIGMETRICS 2005, pp. 97-108, 2005.