This study introduces a predictive, AI-powered auto-scaling framework designed to optimize resource usage in cloud environments, specifically within Amazon Web Services (AWS). Conventional rule-based scaling methods often result in inefficiencies, either wasting resources or degrading performance. To overcome these challenges, this work employs Long Short-Term Memory (LSTM) neural networks that analyze historical performance data collected from AWS CloudWatch. The system forecasts resource demand trends for EC2 and RDS instances and automates scaling actions using the Boto3 SDK. It evaluates multiple metrics—including CPU usage, memory availability, disk I/O, and network traffic—to make accurate, real-time decisions. Operating in a continuous loop, the model updates hourly to adapt to changing workloads. Experimental evaluation confirms that the proposed approach reduces operational costs and enhances performance reliability. This research delivers a scalable, intelligent solution for cloud resource management, suitable for dynamic application environments where responsiveness and efficiency are critical.
Building similarity graph...
Analyzing shared references across papers
Loading...
Sudip Poudel
Pokhara University
Kushal Sharma Marasini
Lokesh Bhatt
Tata Institute of Fundamental Research
Journal of Advanced College of Engineering and Management
Building similarity graph...
Analyzing shared references across papers
Loading...
Poudel et al. (Thu,) studied this question.
synapsesocial.com/papers/68d464ff31b076d99fa64c90 — DOI: https://doi.org/10.3126/jacem.v11i1.84521