RL based OS process Scheduling

Introduction

This project aims to implement a scheduling algorithm for a real-time operating system. We use the MLFQ algorithm as the baseline. In the MLFQ algorithm, there are multiple queues but the quantum of each queue is pre-decided. We aim to address this problem by using an RL agent to control the quantum of each queue.

Code

The Code is divided into multiple files:

• process.py: The process class.
• scheduling_queue.py: The queue class which simulates 1 single queue of the MLFQ algorithm.
• gui.py: The GUI class. To display the schedule generated
• env.py: The OpenAI Gym environment class which simulates the MLFQ algorithm on a randomly sampled instance. It takes the quantum list as the action
• main.py: Takes as input the number of Queues, Agent to train and file for agent and trains that particular algorithm. Currently, only A2C is supported.
• check.py: Used to run the agent given on a test instance.
• check_baseline.py: Used to run the baseline agent on a test instance.

Dependencies

• OpenAI Gym
• Numpy
• PySimpleGUI
• Stable Baselines 3

Output

The output is present in the output directory. The folder name is numberOfQueues_agentType_BoostingValue. rl file contains result for the A2C agent. base file contains the result for the baseline agent. rl_model.zip contains the model weights.