Optimizing Microservices Scalability and Resiliency with Kubernetes

Abstract: This paper proposes a novel hybrid autoscaling system for Kubernetes-based microservices, integrating Horizontal and Vertical Pod Autoscalers with custom metric monitoring and heuristic scheduling strategies (PROP, CONT, UTIL). The proposed approach enhances resource efficiency, fault tolerance, and system responsiveness under variable load conditions. Implementation is validated through real-world deployment using Prometheus, Grafana, and Vegeta, and addresses gaps in existing approaches that rely solely on CPU/memory metrics. Results show improved scalability, resource allocation, and performance. The work contributes a flexible, adaptive model with practical implications for cloud-native application management.