Back in 2009 I was musing about a scheme in which the data of heavy users could be dynamically de-prioritized on the cell level if other users are present. This way the quality of experience, i.e. the data rates available to the the majority of users would not suffer as a result of a few users with intensive bandwidth requirements. On the other hand heavy users would still get high data rates while they were alone in the cell.
At the time I was made aware that Ericsson actually had a flavor of such a feature implemented already but even today, I haven't seen it in use so far. When I recently revisited the topic still thinking it was a good idea I noticed that I had perhaps missed a potential side effect of such an approach:
All cells in UMTS and LTE networks use the same channel and thus the signals of different cells interfere with each other. Heavy activity in one cell reduces the data rates available in neighboring cells as the activity in one cell is seen as noise by devices being served by neighboring cells. So in a scenario where a heavy user is pretty much alone in a cell and is thus allowed to transfer data as fast as possible, a user in a neighboring cell especially at the cell edge will have a much lower data rate available than if the heavy user in the other cell would simply be throttled to a much lower speed.
So to be fair to users in neighboring cells the approach would have to be extended to also take traffic in neighboring cells into account. In practice neighboring cells would have to exchange information about their current load. Not impossible to do, but not trivial to implement either.
Help might come from advanced receivers that include inter cell interference cancellation (ICIC) which would reduce the noise rise issue in the scenario describe above and hence make the exchange of load information between cells superfluous. So even while the "dynamically de-prioritzing heavy users on the cell level" scheme still waits for its use in a live network it remains an interesting "Gedankenexperiment" with a couple of new turns and twists added!
Hmm. Just wondering about this. Usually the BTS has some kind of proportional fair scheduler often with delay sensitivity added on.
Wouldn’t this sort of guarantee that lower bitrate users are not too adversely affected.
Of course it cannot help with bitrate reduction that comes with noise level increase, but there are many techniques for that (IC, FDE, CPC, e-DPCCH boost), which should help there.
Also the question: who should control individual user’s bitrate? Core or RAN?