New round of measuring the accuracy of Tor relays' advertised bandwidth
Hello! You might recall we ran two "speed tests" so far for investigating the accuracy of a relay's advertised bandwidth, one in 2021[1] and another one earlier this year[2]. We gonna pick this test up again this week. However, this time it won't be just a single run but we are rather scheduling the experiment to be on for two weeks and then off for two weeks for the upcoming weeks/months (see: the third item in our network experiments strategy[3] for some rationale). More exactly: we'll run our measurement script for roughly one week, then wait about a week for flooded values leaving the relay descriptors, then 2 weeks off, then starting again... The experiment itself has not changed: we are planning to download several large data streams for a period of about 20 seconds with the goal to get a better estimation of the relay's true capacity. There are more details related to this experiment in particular in an older mailing list thread[1], in case anyone is interested and has not seen those yet. As in previous instances of this "speed test" there is the option for an opt-out. Let us know if any of you doesn't want to participate and we remove your relay(s) from the list to scan. Finally, an up-to-date status about the flooding being on/off (and other experiments scheduled) can be found on our status page[4]. Thanks, Georg [1] https://lists.torproject.org/pipermail/tor-relays/2019-July/017535.htmlff. [2] https://lists.torproject.org/pipermail/tor-relays/2021-May/019649.html [3] https://gitlab.torproject.org/tpo/network-health/team/-/wikis/Network-experi... [4] https://status.torproject.org/affected/network-experiments/
Georg Koppen:
Hello!
You might recall we ran two "speed tests" so far for investigating the accuracy of a relay's advertised bandwidth, one in 2021[1] and another one earlier this year[2].
We gonna pick this test up again this week. However, this time it won't be just a single run but we are rather scheduling the experiment to be on for two weeks and then off for two weeks for the upcoming weeks/months (see: the third item in our network experiments strategy[3] for some rationale). More exactly: we'll run our measurement script for roughly one week, then wait about a week for flooded values leaving the relay descriptors, then 2 weeks off, then starting again...
A small correction here: We'll run the measurement script for roughly *two* weeks, then wait about a week for flooded values leaving the relay descriptors, then 1 week off, then starting again. We figured out that's a better use of our resources compared to the two weeks break in the measurements cycle. Georg [snip]
Georg Koppen:
Georg Koppen:
Hello!
You might recall we ran two "speed tests" so far for investigating the accuracy of a relay's advertised bandwidth, one in 2021[1] and another one earlier this year[2].
We gonna pick this test up again this week. However, this time it won't be just a single run but we are rather scheduling the experiment to be on for two weeks and then off for two weeks for the upcoming weeks/months (see: the third item in our network experiments strategy[3] for some rationale). More exactly: we'll run our measurement script for roughly one week, then wait about a week for flooded values leaving the relay descriptors, then 2 weeks off, then starting again...
A small correction here:
We'll run the measurement script for roughly *two* weeks, then wait about a week for flooded values leaving the relay descriptors, then 1 week off, then starting again.
We figured out that's a better use of our resources compared to the two weeks break in the measurements cycle.
Another cycle of that experiment started a couple of hours ago.[1] We hope to have finally figured out all the bugs in our setup so that we don't see several peaks during our measurement period anymore.[2] Georg [1] https://status.torproject.org/experiments/2021-09-23-relay-speed-test/ [2] See: https://metrics.torproject.org/bandwidth.html?start=2021-05-01&end=2021-09-2... for how the two measurements look like we did this year.
On 8/25/21 12:02 PM, Georg Koppen wrote:
Hello!
You might recall we ran two "speed tests" so far for investigating the accuracy of a relay's advertised bandwidth, one in 2021[1] and another one earlier this year[2].
I do run 2 relays at the same ip address. Do the tests consider that ? -- Toralf
Toralf Förster:
On 8/25/21 12:02 PM, Georg Koppen wrote:
Hello!
You might recall we ran two "speed tests" so far for investigating the accuracy of a relay's advertised bandwidth, one in 2021[1] and another one earlier this year[2].
I do run 2 relays at the same ip address.
Do the tests consider that ?
No. As far as it matters for the test those two relays are independent relays which each get tested as two random relays would get. Georg
-- Toralf _______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
Toralf Förster:
On 9/2/21 9:11 AM, Georg Koppen wrote:
No. As far as it matters for the test those two relays are independent relays which each get tested as two random relays would get.
Except, that both shares the same network card ...
Right, good point. We'll see how that goes. There might be a slight chance that both of them share the same or and overlapping 20s slot. That's a things we should consider when evaluating the experiment data. Georg
-- Toralf _______________________________________________ tor-relays mailing list tor-relays@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-relays
participants (2)
-
Georg Koppen -
Toralf Förster