We provide the Java application specifically designed for task scheduling simulation and Worst-Case Response Time (WCRT) analysis in the context of Linux's Completely Fair Scheduling (CFS). It is important to note that this application is currently under active development and enhancements.

In the real_linux_application, you'll find an application developed for direct WCRT measurement by running tasks on actual Linux systems. For a detailed explanation of the implementation, please refer to the document.

For a comprehensive understanding of the task set and instructions for conducting experiments, please refer to the below.

git clone --recurse-submodules [repository address]

• Java Version: Requires Java 8 or higher.
• Build Tool: Gradle is used for building the application and managing dependencies.
• Refer to build.gradle.kts for detailed dependency information.

• GCC Version: Requires gcc 9.4 or higher.
• Build Tool: Cmake is used for building the application and managing dependencies.
• Refer to CMakeLists.txt for detailed dependency information.
• We tested the application on Raspberry Pi4 which installed Rasbian(Linux Kernel 5.15.92)

pip3 install pandas matplotlib scikit-learn

• We provide the complete set of task sets used in our experiments. (Available in the file)
• To generate new task sets, execute the following command. Note that periods are in microseconds.
• You can modify the number of tasks in a set and the system utilization in main.py.
• For each combination of task numbers and system utilizations, a corresponding number of task sets (num_tasksets) will be generated.

  python3 ./task_generation/main.py --generated_files_save_dir=./generated_taskset_303000 --num_tasksets=100 --min_period=30000 --max_period=3000000

1. Build the application:

   gradle wrapper
   ./gradlew build
   mv ./app/build/libs/run.jar ./run.jar

   *(Optional) The build is successful if executing the command generates "exp_result.csv".

   java -jar run.jar -t=./sample_taskset.json -rd=./ 

2. Configure the experiment settings in the script:
   • num_tasks: List of the number of tasks (default: 4, 8, 12, 16, 20)
   • utilization: List of utilization values (default: 0.2, 0.4, 0.6, 0.8)
   • target_latency: Target latency in microseconds (default: 18,000)
   • min_gran: Minimum granularity in microseconds (default: 2,250)
   • jiffy: Jiffy value in microseconds (default: 1,000)
   • nice_assign: Strategy for assigning nice values (default: GA)
3. Run the script:

   bash ./script/run/proposed_and_simulator/run_exps.sh

   • (Optional) To save experiment time, you can run the program in parallel based on the combination of the number of tasks and utilization settings. Make sure to configure the settings before running the parallel script.

     bash ./script/run/proposed_and_simulator/run_exps_parallel.sh

1. Build the application:

   bash ./script/run/real_linux/build_application.sh

2. Configure the experiment settings in the script:
   • num_tasks: List of the number of tasks (default: 4, 8, 12, 16, 20)
   • utilization: List of utilization values (default: 0.2, 0.4, 0.6, 0.8)
3. Run the application with synthetic task sets:

   sudo bash ./script/run/real_linux/run_exps_with_synthetic_tasksets.sh

   • To specify a task's nice value from -20 to +19, sudo privileges are required.

1. Build the application:

   bash ./script/run/real_linux/build_application.sh

2. Run the application with a real-world application:

   sudo bash ./script/run/real_linux/run_exp_with_real_application.sh

   • The task set originates from satellite onboard software supplied by Thales Alenia Space (TAS), as outlined in the paper.

• Nice Value Assignment : notebook
  • This analysis compares the schedulability estimated from the proposed WCRT analysis method using different nice value assignment strategies.
  • It also examines the time consupmtion for analysis depending on the nice value strategies.
• Overestimation of the Proposed WCRT Analysis : notebook
  • This evaluates the schedulability estimated from the proposed WCRT analysis method against those obtained from the simulator and a real Linux system.
• Suitability of CFS : notebook
  • This analysis assesses the schedulability of task sets under the CFS scheduler with varying task periodicities.
• Comparison of Schedulers : notebook
  • This provides a comparison of worst-case and average response times using CFS and other real-time schedulers.

For more detailed information and instructions, please refer to the script directory in the project.