MSc thesis - Replication package

This repository contains the replication package of my thesis called The Impact of Resource Preloading and Resource Hints on theEnergy Consumption and Performance of Mobile Web Apps.

The repository contains all material required for the replication of the two conducted experiments:

• All subjects and the scripts used during their selection process.
• Our measurement infrastructure consisting of mitmproxy and a slightly modified version of the Android Runner framework.
• All the raw data collected during the execution of the experiments.
• The R notebooks for exploring, summarizing, and analyzing measurement data.

Overview of the replication package

A high level overview of this repository's structure is shown below.

├── android-runner                                  Source code of the Android Runner framework.
├── preloads                                        Contains all resources needed for the preload experiment.
│   ├── AR_config_preloads.json                     AR config file used for this experiment.
│   ├── data_analysis                               R scripts and plots used during the data analysis
│   ├── mitmproxy                                   Our mitmproxy plugin, the used config and all prerecorded responses. 
│   └── raw_data                                    Raw data of the experiment
├── resource_hints                                  Contains all resources needed for the resource hints experiment
│   ├── AR_config_resource_hints.json               AR config file used for this experiment
│   ├── data_analysis                               R scripts and plots used during the data analysis
│   ├── mitmproxy                                   Our mitmproxy plugin, the used config and all prerecorded responses.
│   └── raw_data                                    Raw data of this experiment
└── subject_selection                               All scripts used during our subject selection process.
    ├── scraper                                     Scraper used to mine 3K websites meeting our criteria.
    └── selection                                   
        ├── preloads                                Subjects used for the preload experiment.
        └── resource_hints                          Subjects used for the resource hints experiment.

Replicating the experiment

Our measurement infrastructure is shown below and consists of three components: a mobile device, a RPi and the Monsoon HVPM.

Setup mitmproxy

1. Download and install mitmproxy from here.
2. Configure mitmproxy on your mobile device by installing the required certificate and as shown here. Android requires you to set a lock-screen PIN to add a certificate. Based on your device's resolution and chosen pin you might have to change the android-runner/examples/monsoon/scripts/after_run.py script accordingly.
3. Download the recorded flows from the URLs provided in the .txt files.
4. Make sure mitmproxy uses the correct configuration file. Copy config_preloads.yaml or config_hints.yaml to ~/.mitmproxy and rename it to config.yaml.
5. Start mitmproxy with our plugin by running mitmdump -s preloads_plugin.py or mitmproxy -s resource_hints_plugin.py. Please note that we are using mitmdump instead of mitmproxy!

Setup Android Runner

1. Setup a Python virtual environment using python3 -m venv .venv and install the required dependencies for Android Runner as described here.
2. Setup the Monsoon HVPM following the guide here. Make sure you install the original PyMonsoon package instead of this forked version.
3. Start the experiment by running sudo .venv/bin/python __main__.py AR_config_preloads.json or sudo .venv/bin/python __main__.py AR_config_resource_hints.json. Running the commands with sudo is only required if you are using the Monsoon HVPM.