A. Fairness analysis
It is diffilt to represent fairness quantitatively;
however, some mathematical and conceptual defiitions of
fairnss do exist.
For evaluating the fairnss, we assume heterogeneous
networks with different service requirements. Around 30%
of network users transmit at a fied rate of at least 1 MB/s
(to transmit video streams, for example). Another 40% of
users transmit data, at a constant rate of 120 KB per time
slot. The remaining 30% of users transmit data at a random
rate, with the mean of the distribution being 250 KB per
time slot. As we can see in Fig. 3, the Max Rate policy is
the most unfair algorithm because it allocates system
resources for the users with the strongest channels.
Accordingly, the fairnss index decreases with an increasing
number of users. On the other hand, the round robin
technique is the fairest of the algorithms. In fact, with RR
the BS assigns equal time to each user with a fied
established order.
The PFS algorithm overall has good behavior,
maintaining fairness without degrading system throughput.
ReG algorithms maintain good fairnss until the number of
users on the network exceeds a certain value (10 users in our
simulation). Indeed, with a low number of users, ReG
algorithms have a suffient number of sub-carriers to attend