I do wonder about any changes in the algorithm which drops the cw of a relay [1] by a magnitude within 1 year as seen in [2]?
[1] https://metrics.torproject.org/rs.html#details/1AF72E8906E6C49481A791A6F8F84... [2] https://screenshotscdn.firefoxusercontent.com/images/040958f1-a5be-4630-9551...
Toralf Förster:
I do wonder about any changes in the algorithm which drops the cw of a relay [1] by a magnitude within 1 year as seen in [2]?
[1] https://metrics.torproject.org/rs.html#details/1AF72E8906E6C49481A791A6F8F84... [2] https://screenshotscdn.firefoxusercontent.com/images/040958f1-a5be-4630-9551...
You are looking at cw **fraction**, which is a relative value compared to the rest of the network, which is why I think it isn't the best value to look at when trying to determine "is my relay doing better or worse than usual?"
If your relay is doing great as usual and others add 50 Gbit/s of capacity you might see your cw fraction and exit probability go down.
Your cw fraction and exit probability is decreasing since several month, at the same time the the overall tor network capacity increased:
https://metrics.torproject.org/bandwidth.png?start=2017-08-21&end=2018-0... https://metrics.torproject.org/bandwidth-flags.png?start=2017-08-21&end=...
The better question might be: Did your absolute cw value decrease as well or did it remain static?
This is why I'd like to see absolute CW graphs in addition to the fraction graphs (and even the ratio of them), here is the ticket:
On 8/19/18 11:58 AM, nusenu wrote:
If your relay is doing great as usual and others add 50 Gbit/s of capacity you might see your cw fraction and exit probability go down.
Your cw fraction and exit probability is decreasing since several month, at the same time the the overall tor network capacity increased:
Well, that's indeed the area Id like to understood fully in detail.
The overall bandwidth of the network changed roughly by a factor of 2, whilst my cw fraction value decreased down from 0.245% to 0.032% - which is a factor of 10.
One possible explanation could be that most of the network growth was made at the same AS as where I do reside?
Toralf Förster:
One possible explanation could be that most of the network growth was made at the same AS as where I do reside?
AS is not taken into account for path selection but same /16 IPv4 netblock and same family is.
On 19 Aug 2018, at 22:29, nusenu nusenu-lists@riseup.net wrote:
Toralf Förster:
One possible explanation could be that most of the network growth was made at the same AS as where I do reside?
AS is not taken into account for path selection but same /16 IPv4 netblock and same family is.
Bandwidth authorities do not use these client path selection rules. (Bandwidth authorities select paths based on the Exit flag, and previous bandwidths.)
Client path selection does change relay bandwidth, which then changes: * the residual relay bandwidth available for measurements, and * the scaling applied to raw bandwidth measurements. But these factors tend to cancel each other out.
T
tor-relays@lists.torproject.org
-
nusenu -
teor -
Toralf Förster