Optimal Reconfiguration of Dynamic Software Product Lines Based on Performance-Influence Models

This repository contains supplementary material for the paper Optimal Reconfiguration of Dynamic Software Product Lines Based on Performance-Influence Models. The implementation of our approach for the Adaptive Wireless Sensor Networks case study is called Coala - Context-Aware Topology Control Adaptation.

• coala contains binary that was used to produce the evaluation results. Please have a look at coala_run.sample.bash for further instructions.
• eval_dataset contains the raw simulation output that was used to produce the evaluation plots
• eval_plots contains the plots shown in the paper + some additional plots
• training_dataset contains the raw simulation output that was used for training with SPLConqueror
• traininig_splc contains the training configuration for SPLConqueror (scripts) and the resulting performance-influence models (logs).
  • The file machineLearnSettings.txt specifies the machine-learning settings.
  • We used revision '8ea8ea2203955562205d74aa68452cd69a12afce' of https://github.com/se-passau/SPLConqueror for learning.
  • If you want to reproduce the machine-learning results, you need to replace 'TODO_ADJUST_TO_SUPERSCRIPT_PARENT_FOLDER' with the absolute path to training_splc/scripts in all '*.a' files.