[tor-bugs] #28481 [Core Tor/Tor]: Tor's startup time is getting slower on Android

[Tor Bug Tracker & Wiki]

#28481: Tor's startup time is getting slower on Android
-------------------------------------------------+-------------------------
 Reporter:  akwizgran                            |          Owner:  (none)
     Type:  defect                               |         Status:  new
 Priority:  Medium                               |      Milestone:  Tor:
                                                 |  0.4.0.x-final
Component:  Core Tor/Tor                         |        Version:  Tor:
                                                 |  0.3.4.9
 Severity:  Normal                               |     Resolution:
 Keywords:  android startup performance          |  Actual Points:
  controller                                     |
Parent ID:  #26630                               |         Points:
 Reviewer:                                       |        Sponsor:
                                                 |  Sponsor8-can
-------------------------------------------------+-------------------------

Comment (by akwizgran):

 I'll run some tests with and without cached files, and look into what kind
 of profiling I can do on Android.

 In the meantime here's some more data from the tests I already ran, which
 seems to show there isn't a significant difference in CPU usage across
 versions (or at least, nothing big enough to explain the difference in
 startup time). This was measured on Linux x86_64 with bash's `time`
 builtin. Min and max over 5 runs per version, as before.

 User time:

 ||= Tor version =||= Min =||= Max =||
 ||0.2.9.16||3.55||4.02||
 ||0.2.9.17||3.50||3.88||
 ||0.3.0.1-alpha||3.54||3.79||
 ||0.3.1.1-alpha||3.56||3.87||
 ||0.3.2.1-alpha||3.44||3.92||
 ||0.3.3.1-alpha||3.49||3.86||
 ||0.3.4.1-alpha||3.52||3.78||
 ||0.3.4.8||3.53||3.76||
 ||0.3.4.9||3.57||3.96||
 ||0.3.5.5-alpha||3.43||3.78||

 System time:

 ||= Tor version =||= Min =||= Max =||
 ||0.2.9.16||0.46||0.50||
 ||0.2.9.17||0.39||0.56||
 ||0.3.0.1-alpha||0.44||0.56||
 ||0.3.1.1-alpha||0.38||0.52||
 ||0.3.2.1-alpha||0.44||0.54||
 ||0.3.3.1-alpha||0.40||0.52||
 ||0.3.4.1-alpha||0.40||0.57||
 ||0.3.4.8||0.42||0.57||
 ||0.3.4.9||0.39||0.46||
 ||0.3.5.5-alpha||0.42||0.54||

 Real time:

 ||= Tor version =||= Min =||= Max =||
 ||0.2.9.16||30.37||30.40||
 ||0.2.9.17||30.38||30.39||
 ||0.3.0.1-alpha||30.38||30.39||
 ||0.3.1.1-alpha||30.38||30.39||
 ||0.3.2.1-alpha||30.37||30.39||
 ||0.3.3.1-alpha||30.37||30.39||
 ||0.3.4.1-alpha||30.38||30.39||
 ||0.3.4.8||30.38||30.39||
 ||0.3.4.9||30.38||30.39||
 ||0.3.5.5-alpha||30.37||30.39||

 These are from the same runs as the startup times above, so we'd be
 looking for a difference of about 0.7 seconds to explain the difference in
 startup time. I'm not seeing anything on that scale, or any overall
 pattern. But you're right that the difference in scale between these
 results and the Android results suggests that some resource provided by
 the device is the bottleneck.

 Could it be disk rather that CPU that's the bottleneck resource? That
 would fit with your theory about cached files. I'll get back to you when I
 have those results.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/28481#comment:8>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online