Microsoft Azure Compute

Azure compute is not one product. It is a set of hosting models for application code. You can operate a virtual machine, deploy a web application to a managed platform, run a container, execute a function after an event, or submit parallel jobs.

Start with the workload, not the service name. Ask what must run, how long it runs, what triggers it, and how much infrastructure your team needs to control.

The central tradeoff is control against operational responsibility. Azure Virtual Machines expose the guest operating system and broad configuration choices. Managed platforms such as Azure App Service and Azure Container Apps operate more of the hosting environment. Azure Functions narrows the unit further to event-driven code.

No compute service removes application architecture. You still design identity, network access, state, failure handling, observability, deployment, and cost controls.

Use the execution unit as your first filter

Each service starts and scales a different unit:

• Azure Virtual Machines: a virtual server.
• Virtual Machine Scale Sets: a managed group of virtual machines.
• Azure App Service: a web application, API, mobile back end, or web job.
• Azure Functions: a function invocation started by a trigger.
• Azure Container Apps: a container app revision or a finite container job.
• Azure Kubernetes Service: a Kubernetes workload scheduled onto a managed cluster.
• Azure Container Instances: a container group without a full orchestrator.
• Azure Batch: a pool of compute nodes that runs scheduled parallel jobs.

A legacy server process may require a virtual machine. A web application that fits a supported runtime may fit App Service. A packaged HTTP service may fit Container Apps. A short reaction to a queue message may fit Functions. A Kubernetes platform requirement points toward Azure Kubernetes Service.