AI Newsway
Back to glossary

Kubernetes

Cloud computing

Kubernetes is an open-source platform that runs and manages containers across a cluster of machines. Google released it in 2014 based on its internal experience, and it is now governed by the Cloud Native Computing Foundation; the name is abbreviated K8s. You describe the desired state in YAML, covering how many replicas of an image should run, what resources each needs, and how traffic reaches them, and a control loop continuously reconciles reality with that description. It restarts failed containers, reschedules work when a node dies, and scales replicas against load. The core objects are Pods, Deployments, Services, and Ingress, with Helm packaging them for reuse and operators encoding domain-specific management. Every major cloud offers a managed control plane, such as EKS, GKE, and AKS. AI teams use it to schedule GPU workloads and share expensive accelerators across jobs. The cost is complexity: networking, storage, and permissions each bring their own layer, and a small team can spend more effort on the cluster than on the product. For a handful of services, a managed container platform or a serverless deployment is usually the better trade.