Exploring Multilevel Feedback Queue Combinations and Regression-Based Time Quanta in Scheduling Algorithms

Abstract

Much like day-to-day life, scheduling methods have roles in managing what to do and how much time or resources to invest. Multilevel Feedback Queues (MLFQ) are a widely used algorithm due to its capacity to handle a wide range of task types and execution times. There is increasing interest in exploring dynamic scheduling models that can adapt to live variables solved by machine learning. This study explores an approach to process scheduling by integrating linear regression into the dynamic adjustment of round robin time quanta within MLFQ systems. This paper identifies system resources that can be modelled using linear regression into dynamic round robin, introduces an algorithm using linear regression to optimize system performance in an MLFQ implementation, and explains how dynamic time quanta affect the performance of MLFQ scheduling algorithms. SJF and Round Robin are the most effective combination within the MLFQ system, while FCFS would be the last layer to ensure task completion. Average Waiting Time and Average Turnaround Time are the most appropriate metrics for evaluating MLFQ systems. The integration of linear regression into dynamic time quantum adjustments improves performance by reducing AWT, ATAT, and context switching.