Autoscaling Enabled Intelligent Load Balancing In Cloud Computing

Abstract

Cloud computing environments demand robust and intelligent task management to ensure scalability, efficiency, and cost-effectiveness. This paper introduces an Autoscaling Enabled Intelligent Load Balancing System designed for dynamic cloud environments. The system is divided into four key modules: task classification using XGBoost, resource prediction via XGBoost Regression, optimal load balancing using Particle Swarm Optimization (PSO), and adaptive autoscaling with fuzzy logic. Evaluation of each module demonstrated significant improvements in task distribution accuracy, resource utilization, and system stability. The integration of machine learning and heuristic algorithms ensures a dynamic, responsive, and energy-efficient cloud environment.

Introduction

1. Objective
To address inefficiencies of static load balancing in cloud systems, this project introduces a machine learning and heuristic-based approach for intelligent task scheduling, resource forecasting, and real-time autoscaling. It combines XGBoost, Particle Swarm Optimization (PSO), and Fuzzy Logic for adaptive decision-making.

2. Problem Statement
Traditional load balancing and threshold-based autoscaling suffer from:

• Underutilization of resources

• High task latency and cost

• Lack of task-awareness

The goal is to overcome these with dynamic, predictive, and task-aware mechanisms.

3. Literature Insights
Existing models (e.g., SVM-PSO, LSTM-PSO, Fuzzy Logic) show potential but fall short due to:

• High computational complexity

• Poor scalability

• Abrupt scaling behavior

Conclusion: A hybrid, lightweight, and real-time model is needed.

4. System Architecture
The system consists of four integrated modules:

1. Task Classification – XGBoost classifies tasks (audio, video, text, image)

2. Resource Prediction – XGBoost Regression forecasts CPU and memory usage

3. Load Balancing – PSO assigns tasks to optimal VMs

4. Autoscaling – Fuzzy Logic determines when and how to scale VMs

Technologies used: FastAPI, Streamlit, SQLite/MongoDB

5. Methodology Overview

• Classification: Uses FFT-75 file features; achieves 94.6% accuracy

• Prediction: Inputs include task type, priority, and past usage; RMSE ~0.27 (CPU), R² = 0.89

• Load Balancing: PSO improves response time and VM utilization

• Autoscaling: Fuzzy Logic reduces over/under-provisioning and scales faster than threshold-based methods

6. Datasets

• FFT-75: For task classification (statistical byte-level file features)

• Google Cluster 2019: For resource prediction and autoscaling (real-world traces)

7. Experimental Results

• Classification: F1-score of 93.1%, minimal audio/video confusion

• Resource Forecasting: Accurate with low RMSE and high R²

• Load Balancing:

  • 22% faster response

  • VM utilization up 19%

  • Rejection rate <2%

• Autoscaling:

  • Time to scale down to 3.1s

  • Over-provisioning down by 28%

  • Cost savings ~17%

• UI Feedback: Rated 4.7/5 for usability (Streamlit)

• Stress Test: Successfully handled 5,000 tasks; scaled up to 20 VMs dynamically

Conclusion

This work presents a holistic cloud management system combining ML classification, regression, optimization, and fuzzy logic. The integration of XGBoost, PSO, and fuzzy logic results in an adaptive, stable, and cost-efficient infrastructure. A. Future Enhancements • Integrate LSTM for time-series-based resource forecasting. • Use real-time feedback for adaptive PSO weight tuning. • Add anomaly detection module for security-aware scaling. • Implement container-level orchestration using Kubernetes.

References

[1] Muhammad Adil et al., \"CA-MLBS: Content-aware Machine Learning Based Load Balancing Scheduler in the Cloud Environment,\" Expert Systems, 2023. [2] SalehaAlharthi et al., “Auto-Scaling Techniques in Cloud Computing: Issues and Research Directions,” MDPI, 2024. [3] Sethi, S. et al., \"Efficient Load Balancing in Cloud Using Fuzzy Logic,\" IJCSIT, 2014. [4] Juliet Muchori et al., \"Machine Learning Load Balancing Techniques in Cloud Computing,\" Research Gate, 2022. [5] Bacanin, N. et al., \"Cloud Load Prediction with LSTM & PSO,\" Annals of Operations Research, 2023. [6] RahmatHidayat et al., “Comparative Analysis of Logistic Regression, SVM, XGBoost, and Random Forest Algorithms,” JurnalTeknologiSistemInformasi, 2024. [7] Khan, M.A. et al., \"Load Balancing of Cloud Computing Empowered by Fuzzy Logic,\" SSU Research Journal of Engineering and Technology, 2020. [8] Kumar, P. &Charu, \"Resource Utilization Prediction Using Ensemble Random Forest + LSTM,\" J. King Saud Univ., 2021.

Copyright

Copyright © 2025 Shivam Kumar Choudhary, Surendar U, Yogeshwaran B, Dr. G Santhi. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.