Auto-Scaling and Load Balancing Strategies inKubernetes For High-Availability CloudApplications

Abstract

In cloud computing, virtual infrastructure based on virtual machines has been widely used to support various businesses.
Kubernetes is one of the most iconic of these systems. Kubernetes, K8s in short, is used for managing containers and it is most widely
used. This study examines Kubernetes as a foundational edge orchestration platform within modern cloud–edge computing
environments. It highlights Kubernetes’ role in enabling automated deployment, horizontal scaling, and efficient resource utilisation
across heterogeneous infrastructures, including both x86 and emerging ARM-based platforms. The work classifies Kubernetes-based
edge orchestration into three categories: platform-based solutions that extend Kubernetes without modification, cloud–edge
collaborative architectures that centralize orchestration while offloading execution to the edge, and customized edge-specific solutions
that adapt Kubernetes for constrained fog computing scenarios. Core distributed service mechanisms are discussed in terms of
application-level versus platform-provided service discovery models. The study further analyzes Kubernetes resource management
components, including scheduling, admission control, auto-scaling, load balancing, and health monitoring. Autoscaling
mechanisms—including HPA, VPA, and Cluster Autoscaler—are presented alongside resource and custom metrics pipelines
powered by cAdvisor, Metrics Server, and Prometheus. Finally, load balancing strategies are evaluated, including in-cluster
mechanisms (e.g., kube-proxy), external load balancers (e.g., cloud or MetalLB), and service-mesh-based intelligent routing (e.g.,
Istio, Linkerd). Overall, the study emphasizes Kubernetes’ flexibility and maturity in orchestrating distributed workloads across
cloud–edge systems.