[tor-bugs] #18859 [Core Tor]: A new SOCKS connection should use a pre-built circuit for its first stream

[Tor Bug Tracker & Wiki]

#18859: A new SOCKS connection should use a pre-built circuit for its first stream
-----------------------------+---------------------
 Reporter:  arthuredelstein  |          Owner:
     Type:  defect           |         Status:  new
 Priority:  Medium           |      Milestone:
Component:  Core Tor         |        Version:
 Severity:  Normal           |     Resolution:
 Keywords:                   |  Actual Points:
Parent ID:                   |         Points:
 Reviewer:                   |        Sponsor:
-----------------------------+---------------------
Changes (by nickm):

 * status:  assigned => new


Comment:

 Removing myself as owner. I can't do anything here until we have an
 analysis of what to actually try.

 > what determines the rate at which pre-built circuits are built, and
 could we make a performance improvement if we increased that rate?

 The code is in circuit_predict_and_launch_new, which is called once once
 per second.  We could ramp that up, but it's not a super-cheap function.
 We could also schedule it to get invoked shortly after the first time any
 clean circuit becomes dirty.

 Also, circuit_predict_and_launch_new seems to have quite a few magic
 numbers in it to determine the number of circuits to have for hidden
 services.  Adjusting those might get better results.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/18859#comment:7>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online