Real-Time CPU Burst Time Prediction Approach for Processes in the Computational Grid Using ML

Abstract

Shortest-Job-First (SJF) and other CPU scheduling techniques are achieved by analysing the duration of the CPU bursts in the ready queue processes. Static and dynamic approaches can estimate the time of CPU bursts, although they may not provide accurate or dependable predictions. This research proposes a method based on machine learning (ML) to evaluate the CPU bursts of processes. Feature selection approaches are employed to identify and estimate CPU burst times for grid processes in real-time without spending many computational resources and processing time. The suggested method is tested and evaluated using a grid workload data set known as “GWA-T-4 AuverGrid”, utilising ML approaches such as linear regression and decision trees regression. We conducted an experiment that found a linear correlation between CPU burst strength and process properties to test this. Furthermore, in nearly all cases, we strive to design an algorithm that predicts burst times in real-time with minimal time and space complexity to be implemented in the real world.