การทำคอนเทนเนอร์และการจัดการระบบ

Software behaves differently across machines because of mismatched libraries and configs. Containers package an app with all its dependencies so it runs identically everywhere.

This kills the classic 'it works on my machine' problem.

VMs virtualize an entire operating system and are heavy. Containers share the host kernel and isolate only the application, making them lightweight and fast to start.

You can run many containers where you would run a few VMs.

A container image is a built artifact described by a Dockerfile. It lists the base image, dependencies, and start command.

Built images are stored in a registry (Docker Hub, ECR, GCR). Each image gets a tag like myapp:1.4.2.

In CI/CD, the pipeline builds an image, pushes it to the registry, and deploys it to servers.

Running one container is easy. Running hundreds across many machines, with health checks, scaling, and rolling updates, needs an orchestrator.

Kubernetes is the dominant choice for SaaS at scale.

In Kubernetes, a Pod runs one or more containers. A Deployment declares how many replica Pods you want and keeps them running.

Pods are ephemeral and get new IPs. A Kubernetes Service gives a stable address and load-balances traffic across the matching Pods.

This decouples callers from individual Pod lifecycles.

Kubernetes is declarative: you describe the desired state, and the control loop continuously works to match reality to it.

If a Pod crashes, the controller starts a new one automatically. You manage intent, not individual steps.

Orchestrators perform rolling updates: new-version Pods start while old ones drain, keeping the service available throughout.

If the new version fails health checks, the rollout halts and can roll back automatically.

The Horizontal Pod Autoscaler adds or removes Pod replicas based on CPU, memory, or custom metrics.

This matches capacity to demand automatically, a key cost and reliability win for SaaS.

A typical flow: commit triggers CI, which builds and tests an image, pushes it to a registry, and updates the Kubernetes deployment to the new tag.

This makes deployments repeatable, auditable, and fast.

You learned containerization and orchestration:

• Containers package apps for consistency; lighter than VMs
• Images and registries feed the pipeline
• Kubernetes Deployments, Services, rolling updates, and autoscaling run SaaS at scale

Declarative desired state ties it all into CI/CD.