| Paper Details: | Downloads: 2253 |
| Serial Number: | P1121835636
|
| Title: | Machine Learning based Multilevel Feedback Queue Scheduling
|
| Authors: | Yosef Hasan Fayez Jbara
|
| Corresponding Author: |
Dr. Yosef Jbara
|
| Abstract: | CPU scheduling is a critical component in all types of operating systems. It is a key factor that influences the system efficiency. The efficiency of an operating system strongly relies on the execution behaviour of the scheduling policy adopted. Several different scheduling architectures have been developed. However, multilevel feedback queue scheduling (MLFQ), which is also called a multi-level time sharing policy (MLTP), is the most accepted architecture. Despite its popularity, and because of certain influencing factors, this architecture suffers from certain drawbacks, such as a longer average waiting time (AWT) and higher average turnaround time (ATT), particularly for long processes. This work considers one important influencing factor on the performance of a scheduler, namely the fixed amount of time or time quantum (TQ), given for each queue. and the number of queues defined for this architecture. This work attempts to improve the performance of the scheduler through the employment of machine-learning (ML) based approach (called MLMLFQ) to predict the optimal TQ based on historical data. The developed approach is applied to both real-world and synthetic datasets, and the resulting accuracy of the predicted performance is quantified and compared.
|
| Keywords: | Machine learning, MLFQ, Multilevel queue scheduling, CPU scheduling, Information gain, Relief-F, Random tree, Linear regression, K-NN.
|
| Journal/Conference: | International Journal of Automatic Control and System Engineering
|
| Volume: | 18
|
| Issue: | 2
|
| Submission Date: | 8/30/2018 12:00:00 AM
|
| Review Date: | 10/12/2018 12:00:00 AM
|
| Publishing Date: | 12/31/2018 12:00:00 AM
|
| Article Downloads: | 2253
|
| PDF File Download: |
|
| PPTX File Download: |
|