Which plan?
Consumption Plan
- Default hosting plan with pay-as-you-go model
- Pay only for compute resources when functions are running
- Automatic scaling
Flex Consumption Plan
- High scalability with flexible compute options
- Pay-as-you-go billing
- Dynamic scaling
- Based on configured per-instance concurrency
- Based on number of incoming events
- Cold start reduction
- Specify number of pre-provisioned (always ready) instances
- Automatic scaling based on demand
- Virtual networking support
Premium Plan
- Automatic scaling with enhanced performance
- Uses prewarmed workers for zero-delay execution after idle
- Runs on more powerful instances
- Connects to virtual networks
- Best suited for:
- Continuously or nearly continuously running function apps
- Need for more instance control and multiple function apps on same plan with event-driven scaling
- High number of small executions with high execution bills but low GB seconds
- Additional CPU or memory requirements beyond consumption plans
- Longer execution times than Consumption plan maximum
- Virtual network connectivity requirements
- Custom Linux image deployment needs
Dedicated Plan (App Service Plan)
- Runs at regular App Service plan rates
- Best for long-running scenarios where Durable Functions can't be used
- Best suited for:
- Fully predictable billing requirements or manual scaling needs
- Running multiple web apps and function apps on same plan
- Access to larger compute size options
- Full compute isolation and secure network access (App Service Environment)
- High memory usage and high scale requirements (ASE)
Container Apps
- Fully managed containerised function apps
- Hosted by Azure Container Apps
- Event-driven, serverless, cloud-native architecture
- Run functions alongside microservices, APIs, websites, and workflows
- Best suited for:
- Custom library packaging with function code for line-of-business apps
- Migration from on-premises or legacy apps to cloud-native microservices
- Avoiding Kubernetes cluster overhead and complexity
- High-end processing power requirements with dedicated CPU compute resources