Managed functions that run on triggers for event-driven, short-lived workloads.
Choose Azure Functions for event-driven, short-lived work — responding to HTTP requests, queue messages, blob changes, timers, or event streams — where you want to pay only for execution and scale automatically to zero. It fits glue code, lightweight APIs, and background processing.
Avoid Functions for long-running or always-on workloads, or applications needing fine-grained control over the host. Consumption-plan executions have a time limit; for sustained compute, a web app on App Service or containers on Container Apps is a better fit.
| Hosting model | FaaS |
|---|---|
| Container support | Container-compatible |
| Minimum nodes | Serverless |
| State management | Stateless or stateful (with Durable Functions) |
| Web hosting | Not applicable |
| Autoscaling | Built-in service |
| Load balancer | Integrated |
| Scale limit | 200 instances (Consumption); 1,000 (Flex Consumption) |
| Multiregion | Single region only — external router plus multiple host instances |
| Virtual network integration | Supported (App Service plan, Premium, or Flex Consumption plan) |
| Hybrid connectivity | Supported (Premium or Flex Consumption plan) |
| GPU support | Not supported |
| TLS | Supported |
| Architecture styles | Microservices, event-driven architecture |
| Required skills | Event-driven programming, serverless patterns |
| Operational overhead | Very low — serverless management |
| Best for teams | Developers who build event-driven solutions |
Individual functions are stateless, but Durable Functions add stateful orchestration — chaining, fan-out/fan-in, and long-running workflows — on top of the Functions runtime.
On the Consumption and Flex Consumption plans the platform adds and removes instances automatically based on trigger load, scaling to zero when idle so you pay only for executions.
Yes. The Consumption plan caps execution duration, so workloads that run longer than the limit should use a Premium or Flex Consumption plan, or a different compute service.
This is one candidate from the Azure compute decision guide. The right choice depends on your full requirements — evaluate scaling, cost, and operational fit before committing.
Browse all tools →