Skip to content

Latest commit

 

History

History

fig_05_syscall-heatmap

README.md

Unikraft Syscall Heatmap

Goal

This repository contains the experiment to generate a syscalls heatmap for Unikraft. An excel file which contains the current syscalls status is provided in this repository.

In order to gather syscalls we perform a syscalls analysis of 30 applications by running a dependency analyser tool that we have implemented ourself (see Others section for futher information).

To have the right environment for each application, we have created a docker file per application. These ones are available in the dockerfiles folder. Note that docker must be installed on your machine to perform this experiment.

APPROXIMATIVE EXECUTION TIME=30/45minutes

This experiment requires at least docker client 20 and docker engine 19 otherwise results may be less accurate

How to Use (automatic test)

You can automatically run all the experiment by running the make all command. This command will:

  1. Prepare the environment by executing the ./setup.sh script. Note that this script may take some time...
  2. Run the syscalls experiment by running ./run_all.sh. The results are saved into several json files in the results folder.
  3. Generate the heatmap by running the ./plot.sh script and then the figure is saved into the current repository as fig_05_syscall-heatmap.svg.

How to Use (manual test)

You can also test manually by executing the following procedure.

In order to setup the environment for these experiments, please execute first the ./setup.sh script. This script will setup all the environment to generate the heatmap. Once the setup is done, you can run the dependency analyser by running the ./run_all.sh. Note that the setup takes time since each application must be built as a docker container. The results are saved into several json files in the results folder.

Once all the results are generated, run the ./plot.sh script to generate the heatmap. This one will be generated in the current folder with the following name: fig_05_syscall-heatmap.svg.

Others

Additional information that may be interesting if you want to test a new application.

Gathering syscalls for a new application

If you want to gather yourself the syscalls of an app, you need to download and use our toolchain. Please follow the following steps:

  1. Clone the tools repo with the following command: git clone [email protected]:gaulthiergain/tools.git;
  2. Install GO and setup your GO PATH/ROOT variables;
  3. Run make deps within this folder and then run make;
  4. When you have the tools binary file, you can use it as follow: ./tools --dep -p <program_name> [-o|--options <option>][-t|--testFile <test_file>];
  5. Note that you need to install yourself the application before using the tools. The most easiest solution is to write a docker file.

Example with the nginx web server and a test file:

./tools --dep -p nginx -t ~/testfiles/test_http.json

Please refer to the wiki for further information: https://github.com/gaulthiergain/tools/wiki

Further Information about the heatmap script

The heatmap script allows to generate has also various features:

python3 heatmap.py --help
usage: heatmap.py [-h] [--aggregated-file AGGREGATED_FILE] [--nb-apps NB_APPS]
                  [--folder-to-aggregate FOLDER_TO_AGGREGATE]
                  [--display-syscall-name [DISPLAY_SYSCALL_NAME]]
                  [--save-heatmap [SAVE_HEATMAP]]

There are two different modes:

  1. AGGREGATED_FILE: Use the --aggregated-file argument following by the path of a json file. This one contains an aggregated list of syscalls that were gathered and aggregated by testing 30 applications. The following file "syscalls_sample.json" is provided as sample.
  2. FOLDER_TO_AGGREGATE: Use the --folder-to-aggregate argument following by the path of a folder that contains the json files gathered by the toolchain. These ones must use a particular structure that is defined by the toolchain. An example folder "to_aggregate" is provided as sample.

Note that if you use the --aggregated-file argument, you need to adapt the --nb-apps argument. Its default value is 30.