I think we will have to agree to disagree about this.
Ok, let's focus on the problem first. Conclusions can be made later.
Please help us find out which of these things impact one of your relays.
The only thing from your list, which can have effect - is the properties of other relays and theirs Internet connections. But weight is assigned to single relay, so this differences must be filtered out somehow.
Yes, the Tor network measures it from 4 different locations every few days.
And gives incorrect weight as a result. Stable incorrect result. It is needed to discuss what is a good weight. But that can be done later. My idea is that 1 KiB of load when 1000 KiB of bandwidth available is bad. Just imagine that this 1 MiB/s are really can be used. Of course, not to all network. But if some dial-up node can't use it, this doesn't mean that no one can.
What makes your measurement is more accurate?
1. I was checking this speed with random relays, which mean that high-speed relays was certainly included in the circuits. (this gives 1000000 instead of 58436) 2. I was used obtained value directly to decide if relay is good enough. (this gives 1000000 instead of 5000)
Where are you measuring from?
Ukraine.
Is it close to the relay?
I have made many measurements. Some of them was close to relay, some are not.
How long did it take to do the download?
Here are some results: extendcircuit 0 $BD4354E76929C90B7004FF149A3C52189A3B4634,$A53C46F5B157DD83366D45A8E99A244934A14C46 650 CIRC 10 BUILT $BD4354E76929C90B7004FF149A3C52189A3B4634~Hedgehog,$A53C46F5B157DD83366D45A8E99A244934A14C46~csailmitexit PURPOSE=GENERAL TIME_CREATED=2017-06-14T06:02:21.529447 650 STREAM 13 CLOSED 10 38.229.72.16:443 REASON=DONE $ curl --socks5-hostname localhost:9050 --insecure -O https://38.229.72.16/bwauth.torproject.org/16M % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 16.0M 100 16.0M 0 0 856k 0 0:00:19 0:00:19 --:--:-- 985k
But this relay is close to my location. Let's select another: extendcircuit 0 $38BF40B902ABC23B4E1503BE9131F1A3BF8EBAC5,$A53C46F5B157DD83366D45A8E99A244934A14C46 650 CIRC 11 BUILT $38BF40B902ABC23B4E1503BE9131F1A3BF8EBAC5~bzerorelay1,$A53C46F5B157DD83366D45A8E99A244934A14C46~csailmitexit PURPOSE=GENERAL TIME_CREATED=2017-06-14T06:06:19.551317 650 STREAM 17 CLOSED 11 38.229.72.16:443 REASON=DONE $ curl --socks5-hostname localhost:9050 --insecure -O https://38.229.72.16/bwauth.torproject.org/16M % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 16.0M 100 16.0M 0 0 268k 0 0:01:01 0:01:01 --:--:-- 271k
Of course, it is slower. But still far from utilized 2 KiB/s.
Did you measure from different parts of the world?
I don't have control over multiple locations.
- Choose a relay you control to focus on.
- Send information about the relay's CPU and RAM and configured connection limit.
* CPU: Intel Core i5-4690 * RAM: Team Group Dark-1600, 8 GiB
* RelayBandwidthRate 1 MBytes RelayBandwidthBurst 3 MBytes
- Measure the actual connection limit, bandwidth and latency from the rest of the Tor network. (Or from at least 2 locations in the US and Western Europe.)
I doubt if it is good to test with a loop connections, so I will post speedtest.net results. They give good estimate of my connection properties.
http://www.speedtest.net/my-result/6375679554 DOWNLOAD 94.86Mb/s UPLOAD 95.21Mb/s PING 24 ms SERVER SAINT PETERSBURG
http://www.speedtest.net/my-result/6375676443 DOWNLOAD 89.53Mb/s UPLOAD 94.51Mb/s PING 48 ms SERVER MILAN
http://www.speedtest.net/my-result/6375670157 DOWNLOAD 89.53Mb/s UPLOAD 94.54Mb/s PING 51 ms SERVER DRESDEN
http://www.speedtest.net/my-result/6375655269 DOWNLOAD 85.01Mb/s UPLOAD 24.01Mb/s PING 131 ms SERVER NEW YORK CITY, NY
http://www.speedtest.net/my-result/6375666714 DOWNLOAD 53.70Mb/s UPLOAD 16.19Mb/s PING 205 ms SERVER SAN FRANCISCO, CA
http://www.speedtest.net/my-result/6375685977 DOWNLOAD 24.69Mb/s UPLOAD 13.46Mb/s PING 291 ms SERVER TOKYO
Maybe Atlas graphs also can help: https://s8.hostingkartinok.com/uploads/images/2017/06/caaccae1a967a838871cde... https://s8.hostingkartinok.com/uploads/images/2017/06/60ff981f51002bd177e8c9...
Or:
Change they relay's keys, wait a few weeks, and let us know if the bandwidth measurement is better or worse.
I don't want to lose the state, which reproduces the bug.
If it is better, then the relay was put in a low bucket, and was stuck in that bucket. This can happen at random, or if the relay was slow in the past.
My relay was never slow. Possibility of such random stuck is a thing, which is needs to be eliminated.
-- Vort