|
Hi, I think that in this case, every existing connection and
transfers act as "best effort" (it means that used bandwidth for
some connection aren't absolutely reserved), the new transfer is
making a place by reducing just a little bit the others
connection's available speed. If 50 active connection are using 1 MB/s each at the same time, with a total amount of 50 MB/s which is the maximum of a given server for example, then a new connection trying to get 1 MB/s will make the 50 Megabits divided by 51, then the usable bandwidth for all connection becomes 0,98 MB/s in the place of 1 MB/s When your relay is very overused and severe congestion occurs,
then the bandwidth measurement will be very poor, this will reduce
the weight of your relay in the consensus in order to reduce the
congestion. The algorithm is designed to give the ideal consensus
weight for your relay in order to be as close as possible of your
maximum available bandwidth, without placing too much users on it
(with too much users on it, the total used bandwidth will be at
the maximum, but each user get a painful browsing experience
because the available bandwidth per user becomes too weak) Bye ! Julien PS : if the advertised bandwidth is incorrect then congestion will occur and the consensus weight of your relay will be decreased because of poor performance during measurements, that's how possible cheating/disturbance is avoided in this case, I think. I think there is some documentation on this subject, by searching
if you have some time and want to know more on this subject. Le 20/02/2017 à 18:39, Marcobr a
écrit :
|
_______________________________________________ tor-relays mailing list tor-relays@xxxxxxxxxxxxxxxxxxxx https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays