commit 4448f5fc0de60763d42e724036bda8388435af91 Author: Karsten Loesing karsten.loesing@gmx.net Date: Fri Nov 22 09:47:20 2019 +0100

Improve runtime performance of hidserv module.

Fixes #25924. --- CHANGELOG.md | 5 ++ .../metrics/stats/hidserv/Extrapolator.java | 82 ++++++++++------------ 2 files changed, 43 insertions(+), 44 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md index fe74cf5..797d54b 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,5 +1,10 @@ # Changes in version 1.?.? - 2019-1?-??

+ * Medium changes + - Improve runtime performance of the hidserv module by storing + extrapolated statistics even if computed network fractions are + zero, to avoid re-processing these statistics over and over. +

# Changes in version 1.3.0 - 2019-11-09

diff --git a/src/main/java/org/torproject/metrics/stats/hidserv/Extrapolator.java b/src/main/java/org/torproject/metrics/stats/hidserv/Extrapolator.java index 0dcc638..7a44425 100644 --- a/src/main/java/org/torproject/metrics/stats/hidserv/Extrapolator.java +++ b/src/main/java/org/torproject/metrics/stats/hidserv/Extrapolator.java @@ -197,52 +197,46 @@ public class Extrapolator { } }

- /* If we don't know a single consensus with valid-after time in - * the statistics interval, skip this stat. */ - if (consensuses == 0) { - continue; + /* Compute means of network fractions, or assume 0.0 if we don't + * know a single consensus with valid-after time in the statistics + * interval. */ + double fractionRendRelayedCells = consensuses == 0 ? 0.0 + : sumFractionRendRelayedCells / consensuses; + double fractionDirOnionsSeen = consensuses == 0 ? 0.0 + : sumFractionDirOnionsSeen / consensuses; + + /* Extrapolate network totals. If we don't know a single + * consensus, store an empty statistic anyway to avoid processing + * these reported statistics over and over. */ + ExtrapolatedHidServStats extrapolated = + new ExtrapolatedHidServStats( + statsDateMillis, fingerprint); + if (fractionRendRelayedCells > 0.0) { + extrapolated.setFractionRendRelayedCells( + fractionRendRelayedCells); + /* Extrapolating cells on rendezvous circuits is as easy as + * dividing the reported number by the computed network + * fraction. */ + double extrapolatedRendRelayedCells = + stats.getRendRelayedCells() / fractionRendRelayedCells; + extrapolated.setExtrapolatedRendRelayedCells( + extrapolatedRendRelayedCells); } - - /* Compute means of network fractions. */ - double fractionRendRelayedCells = - sumFractionRendRelayedCells / consensuses; - double fractionDirOnionsSeen = - sumFractionDirOnionsSeen / consensuses; - - /* If at least one fraction is positive, extrapolate network - * totals. */ - if (fractionRendRelayedCells > 0.0 - || fractionDirOnionsSeen > 0.0) { - ExtrapolatedHidServStats extrapolated = - new ExtrapolatedHidServStats( - statsDateMillis, fingerprint); - if (fractionRendRelayedCells > 0.0) { - extrapolated.setFractionRendRelayedCells( - fractionRendRelayedCells); - /* Extrapolating cells on rendezvous circuits is as easy as - * dividing the reported number by the computed network - * fraction. */ - double extrapolatedRendRelayedCells = - stats.getRendRelayedCells() / fractionRendRelayedCells; - extrapolated.setExtrapolatedRendRelayedCells( - extrapolatedRendRelayedCells); - } - if (fractionDirOnionsSeen > 0.0) { - extrapolated.setFractionDirOnionsSeen( - fractionDirOnionsSeen); - /* Extrapolating reported unique .onion addresses to the - * total number in the network is more difficult. In short, - * each descriptor is stored to 12 (likely) different - * directories, so we'll have to divide the reported number by - * 12 and then by the computed network fraction of this - * directory. */ - double extrapolatedDirOnionsSeen = - stats.getDirOnionsSeen() / (12.0 * fractionDirOnionsSeen); - extrapolated.setExtrapolatedDirOnionsSeen( - extrapolatedDirOnionsSeen); - } - extrapolatedStats.add(extrapolated); + if (fractionDirOnionsSeen > 0.0) { + extrapolated.setFractionDirOnionsSeen( + fractionDirOnionsSeen); + /* Extrapolating reported unique .onion addresses to the + * total number in the network is more difficult. In short, + * each descriptor is stored to 12 (likely) different + * directories, so we'll have to divide the reported number by + * 12 and then by the computed network fraction of this + * directory. */ + double extrapolatedDirOnionsSeen = + stats.getDirOnionsSeen() / (12.0 * fractionDirOnionsSeen); + extrapolated.setExtrapolatedDirOnionsSeen( + extrapolatedDirOnionsSeen); } + extrapolatedStats.add(extrapolated); } }