Quote:
|
Originally Posted by dragon
IN which case are NTL going to use some type of QOS/Load balanceing? so that say 3 - 4 users down saturate the whole pipe making it slow for everyone else.
I'm not sure if what im being clear enough but what im trying to say is its better that everyone gets say 300kb/s downloads than someone getting 1mbyte/s and everyone else getting stuck with 10kb/s or something during a busy period.
then of course when there is less load on the pipe everyones speed increases equally...
|
It depends whether the pipe is like a motorway where the traffic eventually fills the pipe to saturation or whether the traffic is already at maxumum, or not likely to increase dramatically.
See 10 users all trying to download 900kbs on a 300kbs line, each will take 3 seconds to download it so in total all 10 downloads will occur concurrently and take 3 seconds
the same 10 users on 10Mb, only 4 can do that at a time, but now the 900k file will take 0.1 of a second, then the next four can do that taking another 0.1 of a second and finally the last 2 at 0.1 of a second. So the 2 that were last in this theoretical queue had to wait 0.3 of a second 0.2 due to contention and 0.1 for the download, however, that is still 2.7 seconds quicker than if they were all on 300k.
Also servers often are not connected to the internet by 10Mb pipes, but maybe 2Mb, so if these 10 users on 10Mb all want to download something at the same time, but find themselves connected to 2Mb servers, or servers that will only offer 2Mb to each user then between them they can only utilise 20Mb of the 38Mb for the downloading.
I'm sure the cablecos have better modeling / stats than the examples I have just given, but hopefully that should show why things should not automatically get slowere for everyone and that the local UBR should not grind to a halt.