Root Cause Analysis Calibration Data

This dataset contains three parts:

sizeEvolution.tar: Measurement data of relative standard deviations of generated call trees

RCA strategy data: strategy-COMPLETE.tar, strategy-UNTIL_SOURCE_CHANGE.tar, strategy-LEVELWISE.tar: Measurement data of root cause analysis of example performance differences of 1.0%, 2.0%, 3.0% and 5.0% for three RCA strategies

extract.sh: Script for unpacking the RCA strategy data

RCA-strategy-results.tar.xz: The result data of the analysis

UNTIL_SOURCE_CHANGE_2_1.010.pdf and UNTIL_SOURCE_CHANGE_8_1.010.pdf: Two example files, showing how the F1 score declines with increasing tree depth.

Thereby, the measurability of root causes of performance changes based on the effect size and measurement strategy and configuration can be estimated.

To execute the analysis of the data, execute the following steps:

Download everything to the folder $DATA (eg. ~/Downloads/data/).

Extract the data by running ./extract.sh in this folder.

Analyse the data by running scripts/analysis_strategy/localStrategyComparison.sh $DATA from the precision-experiments-rca repository (link will be provided upon acceptance), where $DATA is the folder of the data. This will yield pdf result files like given in RCA-strategy-results.tar.xz.

To execute the measurements, use the precision-experiments-rca repository (link will be provided upon acceptance).

The measurement data were obtained by execution of https://github.com/DaGeRe/precision-experiments-rca/ on Java 1.8.0_312 running on Ubuntu 20.04 and on i7-6700 @ 3.40 GHz.