Optimizing Round Robin Scheduling with DBSCAN Clustering and Machine Learning

In the intricate realm of operating systems, scheduling algorithms play a pivotal role in resource allocation and process completion, directly impacting overall system performance. The quest for an efficient and optimized scheduling algorithm is perpetual in the pursuit of optimal operating system u...

Full description

Saved in:
Bibliographic Details
Published in:2024 IEEE International Conference on Signal Processing, Information, Communication and Systems (SPICSCON) pp. 1 - 6
Main Authors: Hossain, Md Mehrab, Biva, Nanziba Khan, Nahid, Nafiz, Rifat, Ariful Islam, Islam, Ashraful
Format: Conference Proceeding
Language:English
Published: IEEE 01.11.2024
Subjects:
Clustering
DBSCAN
Dynamic scheduling
Feature extraction
Long short term memory
LSTM
Machine Learning
Machine learning algorithms
Operating systems
Round robin
Schedules
Signal processing algorithms
Switches
System performance
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In the intricate realm of operating systems, scheduling algorithms play a pivotal role in resource allocation and process completion, directly impacting overall system performance. The quest for an efficient and optimized scheduling algorithm is perpetual in the pursuit of optimal operating system utilization. The Round Robin algorithm, known for its robust time-sharing methodology, emerges as a stalwart in the scheduling context. Its versatility spans various domains, making it a preferred choice in both general-purpose computing environments and real-time systems. This study aims to elevate the performance of the Round Robin algorithm in terms of average waiting time and number of context switches without altering its core algorithmic structure. A novel approach is proposed, utilizing DBSCAN to group related processes and determine a more accurate and efficient time quantum. Departing from the conventional practice of using a single time quantum for the entire schedule, we divide the schedule into groups. The optimal time quantum for each group is derived through machine learning and deep learning algorithms, leveraging rigorous features extracted from the schedule. Notably, the LSTM model emerges as the top performer, achieving an impressive 97% accuracy. The proposed modified version consistently outshines its traditional counterpart. In 92% of cases, the modified version demonstrates superior performance in average waiting time while achieving a 96.5% improvement in context switching. Considering both metrics, the modified version showcases a notable enhancement in 87.8% of cases. This holistic assessment unveils a 5% reduction in average waiting time and a substantial 15% decrease in context switching, signifying a meaningful advancement in overall system performance.
DOI:10.1109/SPICSCON64195.2024.10941280