Produce visualisations for S61 Guard experiments

This ticket is to establish the visualisations to add to Onionperf to support S61 experiments, for example guard sets over time and guard events and rotation over time. The aim is to investigate guard usage in general (and ensure it is happening properly).

/cc @asn @mikeperry

added Sponsor 61 - FINISHED label

mentioned in commit acute/onionperf@f8be5ee2

mentioned in merge request !10 (merged)

changed milestone to %Sponsor 61 - Making the Tor network faster & more reliable for users in Internet-repressive places

added S61-O1-Maybe - FINISHED label

mentioned in commit acute/onionperf@a1efcca9

mentioned in commit acute/onionperf@e4e8c056

mentioned in commit acute/onionperf@9fb6d22c

mentioned in issue #33421 (closed)

mentioned in issue #40015 (closed)

mentioned in commit acute/onionperf@4165e3bf

added Next label

Additionally, as part of this, we should verify that onionperf is not ever recording measurements from between BUILDTIMEOUT_SET RESET and BUILDTIMEOUT_SET COMPUTED, after we change guards. That is, if we have not already verified this.

@asn I'm picking this up from https://gitlab.torproject.org/tpo/metrics/onionperf/-/issues/33421#note_2733112

I believe you wanted to have a look at the control port log file - https://allen.netstat.org.uk/pub/onionperf_2021-04-06_23:59:59.torctl.log - to see whether it matched up with the actual guards tor used, as my graph indicated this was not the case (the brown dots):

Let me know if you need anything else for this.

On my end, I'd like to know if we want to include this graph in the PDF produced by Onionperf and if you had any other ideas about graphs that might be useful.

marked this issue as related to #33421 (closed)

@acute are you currently working on this issue? We were discussing with @mikeperry how to compare results from shadow simulations and maybe an idea would be to include some additional graph to onionperf.

@hiro I am coming to the end of my contract, but have allocated some hours for this work! What graphs do you have in mind?

@mikeperry what do you think would be ideal to have in onionperf that we don't need/don't have yet in metrics-web?

@mikeperry what do you think would be ideal to have in onionperf that we don't need/don't have yet in metrics-web?

@hiro - the main two things we need in metrics-web is the throughput patch and the filtering patch. These two tickets: statistics#40005 (moved) and website#40005 (closed)

We should not need anything in onionperf itself for this. That work is already done.

closed