[metrics-bugs] #33974 [Metrics/Onionperf]: Update OnionPerf to TGen 1.0.0

[Tor Bug Tracker & Wiki]

#33974: Update OnionPerf to TGen 1.0.0
-------------------------------+------------------------------
 Reporter:  karsten            |          Owner:  karsten
     Type:  defect             |         Status:  needs_review
 Priority:  Medium             |      Milestone:
Component:  Metrics/Onionperf  |        Version:
 Severity:  Normal             |     Resolution:
 Keywords:                     |  Actual Points:
Parent ID:                     |         Points:
 Reviewer:                     |        Sponsor:
-------------------------------+------------------------------

Comment (by karsten):

 Replying to [comment:4 robgjansen]:
 > Also, BTW, I've always thought the torperf timeouts were crazy high and
 not that realistic. In my recent research using Shadow, I found that the
 following values are more realistic. However they will likely result in
 more "timeouts" appearing in the OnionPerf output and this would introduce
 an inconsistency with all of the previous data from torperf, so I'm not
 sure if it makes sense to change the timeouts now or not.
 >
 > {{{
 > g.add_node("stream50k", sendsize="1 KiB", recvsize="50 KiB", timeout="15
 seconds", stallout="0 seconds")
 > g.add_node("stream1m", sendsize="1 KiB", recvsize="1 MiB", timeout="60
 seconds", stallout="0 seconds")
 > g.add_node("stream5m", sendsize="1 KiB", recvsize="5 MiB", timeout="120
 seconds", stallout="0 seconds")
 > }}}
 >
 > (I.e., disable stallouts and use much shorter absolute timeouts only.)
 >
 > FWIW, this is what I now use in Shadow, and I think it would give us a
 better sense of timeouts as users actually experience them.

 The original timeouts used by Torperf were designed so that we use all
 available time between starting new runs. In Torperf, the three file sizes
 were downloaded in parallel, with different start times. For example, a 5
 MiB download would start once per hour, so there was no difference between
 timing out after 120 seconds or leaving it running for up 59 minutes to
 see what happens. But you're right that it's not very realistic.

 With OnionPerf this is a bit different, with new runs starting every 5
 minutes, picking a file size at random. So, yes, I agree that we should
 make that change.

 Let me try to get the recent OnionPerf with the recent TGen running first
 before making more changes. I might move the timeouts/stallouts change to
 another ticket then.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/33974#comment:6>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online