[tor-bugs] #28588 [Core Tor/sbws]: SBWS 'bw_torflow_scale' does not appear to honor relay MaxAdvertisedBandwidth

[Tor Bug Tracker & Wiki]

#28588: SBWS 'bw_torflow_scale' does not appear to honor relay
MaxAdvertisedBandwidth
---------------------------+-------------------------------------
 Reporter:  starlight      |          Owner:  (none)
     Type:  defect         |         Status:  new
 Priority:  Medium         |      Milestone:  sbws 1.0 (MVP must)
Component:  Core Tor/sbws  |        Version:  sbws: 1.0.0
 Severity:  Normal         |     Resolution:
 Keywords:                 |  Actual Points:
Parent ID:                 |         Points:
 Reviewer:                 |        Sponsor:
---------------------------+-------------------------------------

Comment (by teor):

 Replying to [comment:6 juga]:
 > As far as i understand, torflow tracks new consensuses, instead sbws
 gets it at some intervals.
 > For this reason, to obtain the observed bandwidth from all the results
 of a relay, we implemented to obtain both from the mean of those results
 or the last. By default, it's currently using the mean.
 > I think it should probably default to the last.

 Yes, torflow uses the last, so we should use the last. See #28596.

 We can change this decision in 1.1 if we think #27788 and #27789 are good
 ideas.

 > Since we need to also get the average bandwidth and the burst bandwidth,
 should these values be obtained from the last or the mean?

 Torflow uses the latest descriptor, so sbws should too.

 Replying to [comment:7 juga]:
 > Replying to [comment:5 teor]:
 > > We should also remove the comment about MaxAdvertisedBandwidth in the
 sbws scaling section.

 We should put this comment where we calculate the minimum.

 > Shouldn't we also use the minimum of all the values in the descriptor
 instead of of the descriptor bandwidth average?.

 Yes, we should do what torflow does: take the minimum of all the values on
 the bandwidth line in the descriptor.

 > I think that could make the linear method look closer to the torflow's
 one.

 The linear method is not important for sbws 1.0, so you can fix it if you
 want, or leave it.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/28588#comment:8>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online