Monitoring-Driven Auto-Scaling for Cost-Efficient Java Cloud Applications

This research paper studies how monitoring-driven auto-scaling improves cost efficiency in Java-based cloud applications.

Cloud platforms provide automatic scaling, but many systems rely only on CPU usage to trigger scaling decisions. In Java applications, this approach is insufficient because JVM behavior (such as memory usage and garbage collection) directly affects performance. Ignoring these internal metrics often leads to over-provisioning (higher cost) or under-provisioning (poor performance).

The paper analyzes key monitoring metrics including CPU utilization, JVM memory usage, and response time. It examines horizontal and vertical scaling strategies and explains how combining multiple metrics leads to more accurate scaling decisions.

The study concludes that integrating real-time monitoring with auto-scaling policies helps reduce unnecessary resource usage while maintaining stable performance. It also suggests future improvements through predictive and adaptive scaling techniques.