[tor-relays] Inconsistent BW measurements of unused relay

[teor]

> On 12 Dec. 2016, at 00:08, Rana <ranaventures at gmail.com> wrote:
> 
>>> On 10 Dec. 2016, at 07:12, Rana <ranaventures at gmail.com> wrote:
>>> 
>>> My relay remains severely under-used. One thing that bothers me are inconsistent bandwidth measurements. Here they are: 
>>> Atlas “advertised” (which is actually supposed to be “measured”?:   100 KB/s =                                 ~ 800,000 bit/s
>> 
>> This is the minimum of:
>> * the bandwidth rate,
>> * the bandwidth burst, and
>> * the observed bandwidth (the maximum bandwidth your relay has recently
>> sustained over a 10 second period).
>> * the consensus weight, converted to a bandwidth figure (I think?).
>> 
>> If you hover over the figure in atlas, it will break it down for you.
> 
> Thanks for the tip. It says it is the actually measured bandwidth

This is the "Advertised Bandwidth".

It is the minimum of 3 (or 4) different bandwidths your relay
advertises it can handle.

Some of these are measured at your relay as a 10-second sustained
maximum. Others are taken from the configuration.

>>> “I have sent” reported in Tor log: on the average pretty stable 17 mbytes every 6 hours =               ~ 200 bit/s
>> 
>> This is what your relay has actually sent.
> 
> Totally inconsistent with the rest
> 
>>> Atlas graphs:  1 Kbytes/s  on the average                                                                                                      ~ 8,000 bit/s
>> 
>> This is the value that your relay reports it has sent.
>> It is rounded and averaged to preserve client privacy.
> 
> This is not consistent with the 200 bit/s figure. Do you mean to say that Atlas rounds 200 bps on the average to 8000 bps on the average?

This is the "Bandwidth History".

These are measured at your relay over 10-second periods, then averaged
over a few hours, then rounded to the nearest kilobyte.

Since a relay's reported bandwidth history is rounded to kilobytes per
second, you will only see figures in multiples of 8192 bps.

>>> Consensus BW: 26 =                                                                                                                                           ~ 26,000 bit/s
> 
>> This is the low-median of the measurements of the 5 bandwidth authorities. It is a dimensionless figure that only makes sense when compared with other relay consensus weights.
> 
> Can't comment because I have no idea what the formula is, therefore this figure is meaningless to me.

This is the "Consensus Weight", based on the measured bandwidth.

Your relay is measured at random times and in random pairings with other
relays, from 5 different tor clients around the US and western Europe.

Then these measurements are averaged, scaled according to the ratio
between your relay's reported bandwidth and measured bandwidth, and
then the low-median (3rd highest out of 5) is used as the measurement.

If that doesn't make sense, don't worry about the details.

>>> Average upload bw reported by arm: 100 kb/s =                                                                                         ~ 100,000 bit/s
> 
>> I suspect this is actually kilobytes, and is the same as the atlas figure. (They use the same backend library.)
> 
> This would not be consistent at all with actual reported upload of 17 mbytes in 6 hours which as I said is pretty constant. The ~100 Kb/s average bit rate figure reported by arm lingers for HOURS. This rate,s 17 MB would have been sent in THREE MINUTES. If the rate were 100 kbyte/s as you suggest then it would take the relay 22 SECONDS to send what it claims it is sending in 6 hours.

This is the "Advertised Bandwidth".
It is the same one displayed by Atlas.
But it should probably have a capital "B" in "Bps".

>>> Makes zero sense.
> 
> Still doesn't. Why do Tor and Tor-related projects such as arm  publish all these TOTALLY inconsiostent figures? If they want to confuse the adversaries I doubt that it worked, but they sure as hell were highly successful in confusing me :)

It seems to me that you want everything containing the word "bandwidth"
to have the same value. But they are measured over different times and
at different places, so they are never going to be the same.

Have you tried checking the bandwidth of your connection to servers
located in 5 different countries?

I bet you get 5 different answers.

T

-- 
Tim Wilson-Brown (teor)

teor2345 at gmail dot com
PGP C855 6CED 5D90 A0C5 29F6 4D43 450C BA7F 968F 094B
ricochet:ekmygaiu4rzgsk6n
xmpp: teor at torproject dot org
------------------------------------------------------------------------