[tor-relays] Consensus Weight calculation

Vort vvort at yandex.ru
Wed Jun 14 06:49:57 UTC 2017


> 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.


> 1. Choose a relay you control to focus on.
> 2. 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


> 3. 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/caaccae1a967a838871cdea5739e9b7d.png
https://s8.hostingkartinok.com/uploads/images/2017/06/60ff981f51002bd177e8c991748205cf.png


> 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



More information about the tor-relays mailing list