[metrics-bugs] #34257 [Metrics/Onionperf]: Analyze unusual distribution of time to extend to first hop in circuit

[Tor Bug Tracker & Wiki]

#34257: Analyze unusual distribution of time to extend to first hop in circuit
-------------------------------+--------------------------------
 Reporter:  karsten            |          Owner:  metrics-team
     Type:  defect             |         Status:  new
 Priority:  Medium             |      Milestone:
Component:  Metrics/Onionperf  |        Version:
 Severity:  Normal             |     Resolution:
 Keywords:                     |  Actual Points:
Parent ID:                     |         Points:
 Reviewer:                     |        Sponsor:  Sponsor59-must
-------------------------------+--------------------------------

Comment (by dennis.jackson):

 I think arma's 1st hypothesis is exactly right. I ran Arthur's Exit
 Scanning  [https://github.com/arthuredelstein/tor_dns_survey script] over
 the weekend which measures relay latency with two strategies:

  * Fix a guard and build a 2-hop circuit for each exit.
  * Fix an exit and build a 2-hop circuit for each guard.

 The distribution of round trip times is roughly equivalent for both,
 however the circuit construction times are much longer in the latter case
 and roughly inline with what !OnionPerf reports. This seems to confirm the
 theory as the first case we can reuse the existing connection, but the
 latter case cannot.

 I thee 'hidden' round trips prior to the first extension does seem to
 account for all of the skew and the scaling effect. It also leads to the
 question of what the right mode for !OnionPerf is? A different sampling
 strategy would report much better first hop latency.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/34257#comment:16>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online