Azure Kubernetes Service (AKS) — Containers at Scale

“Welcome back. Today we're diving into containers and Kubernetes — the backbone of modern cloud-native architecture. AKS is Azure's fully managed Kubernetes service. Microsoft handles the control plane — upgrades, patches, high availability — so you focus on your applications. If you're building microservices or need to run many containers at scale, this is the video for you.”

“A virtual machine virtualizes hardware — it includes a full copy of an operating system. A container is much lighter — it shares the host OS kernel and packages only the application and its dependencies. Containers start in seconds versus minutes for VMs, use a fraction of the memory, and are completely portable. A Docker image you build on your laptop will run identically in AKS production. This portability and consistency is why containers dominate modern application deployment.”

“When you have dozens or hundreds of containers, you need an orchestrator to manage them. Kubernetes is the industry standard. You tell Kubernetes what you want — 'run 3 replicas of my web app, keep them running, expose them on port 80' — and Kubernetes makes it happen and keeps it that way. If a container crashes, Kubernetes restarts it. If a node fails, Kubernetes reschedules containers on healthy nodes. It's the operating system for your cloud-native applications.”

“Let me walk through the key Kubernetes concepts in the AKS context. Your cluster has a control plane — managed by Azure, you don't touch it — and worker nodes, which are Azure VMs you pay for. A Pod is the smallest unit of deployment — usually one container. A Deployment manages multiple identical pods and ensures your desired replica count is always running. A Service gives your pods a stable endpoint — either an internal cluster IP or an Azure Load Balancer for external traffic.”

“Running your own Kubernetes cluster is complex and expensive — you need to manage etcd, the API server, controller manager, and scheduler, then patch and upgrade them. AKS eliminates all of that. The control plane is fully managed by Microsoft at no cost — you only pay for the worker nodes. AKS integrates with Azure Monitor for metrics and logs, Azure Container Registry for private images, and Entra ID for cluster authentication. And the Cluster Autoscaler automatically adds and removes nodes based on demand.”

“Deploying to AKS is declarative — you write a YAML manifest describing what you want. The Deployment manifest specifies your container image, how many replicas, CPU and memory limits. The Service manifest exposes those pods. You apply both with kubectl apply. Kubernetes reads the manifests, compares to current state, and reconciles any differences. This declarative model means you can store your entire deployment configuration in Git — known as GitOps.”

“AKS gives you multiple layers of scaling. You can manually scale the number of pod replicas with kubectl. The Horizontal Pod Autoscaler watches CPU and memory metrics and adjusts replica count automatically. The Cluster Autoscaler watches for pods that can't be scheduled because nodes are full — when that happens, it adds a new node. KEDA extends this with event-driven scaling — for example, scale your consumer pods based on the number of messages in a Service Bus queue.”

“AKS networking uses Azure CNI, giving every pod a real VNet IP address — they're fully integrated with your Azure network. An Ingress Controller routes external HTTP/HTTPS traffic to the right service based on URL paths or hostnames. For stateful applications that need persistent storage, you mount Azure Disks or Azure Files as persistent volumes. Azure Container Registry stores your private container images and integrates directly with AKS without any credential management.”

“Let's create a real Kubernetes cluster. I'll use the Azure Portal to spin up a two-node AKS cluster, then connect to it using kubectl in Cloud Shell. I'll deploy a sample web application, expose it with a LoadBalancer service to get a public IP, and then scale the deployment to three replicas. The whole thing takes about 10 minutes — this is how fast you can get Kubernetes running on Azure.”

“AKS and containers are how the industry builds scalable, resilient applications today. You now understand the fundamentals — from containers to clusters to autoscaling. Next we shift to data — Azure SQL Database, Microsoft's fully managed relational database service. Databases are the heart of most applications, and Azure makes them far easier to run than managing SQL Server yourself. See you in the next one.”