|
|
@@ -7,16 +7,19 @@
|
|
|
Timeout learning to use the weighted average of the top N=3 modes. This
|
|
|
should improve the timeout calculation in some cases, and prevent
|
|
|
extremely high timeout values. Bug 1335, bugfix on 0.2.2.2-alpha.
|
|
|
- - Implement a filtering step to recompute synthetic build times every time
|
|
|
- the timeout changes. Additionally, place a lower cap on synthetic build
|
|
|
- times, and allow this cap to be controlled by the consensus. This should
|
|
|
- also improve the build time calculations, and should eliminate a case
|
|
|
- where Tor was allocating an excessive amount of temporary memory during
|
|
|
- timeout calculation. Bugs 1335 and 1245, bugfix on 0.2.2.2-alpha.
|
|
|
+ - Alter calculation of Pareto distribution parameter 'Alpha' to use a
|
|
|
+ right censored distribution model. This allows us to remove the synthetic
|
|
|
+ timeout generation and instead calculate build timeouts using truncated
|
|
|
+ times. Bugs 1245+1335, bugfix on 0.2.2.2-alpha.
|
|
|
+ - Keep circuits open (but do not use them) after the circuit timeout for
|
|
|
+ up until the time corresponding to the 95th percentile on the Pareto CDF
|
|
|
+ or 60 seconds, whichever is greater. This is done to provide better data
|
|
|
+ for the new Pareto model. This percentile can be controlled by the
|
|
|
+ consensus.
|
|
|
|
|
|
o Minor bugfixes:
|
|
|
- Eliminate a case where a circuit build time warning was displayed after
|
|
|
- network connectivity resumed.
|
|
|
+ network connectivity resumed. Bugfix on 0.2.2.2-alpha.
|
|
|
|
|
|
o Minor features:
|
|
|
- Add a TIMEOUT_RATE keyword to the BUILDTIMEOUT_SET control port event,
|
Mike Perry