A CloudRadial AutomationAI runner hosts its Function Apps on one of two Azure hosting plans: Flex Consumption or Premium. This article explains the difference, the cost and cold-start trade-off, and how to choose. It is for the technician installing or operating a runner.
- Flex Consumption
- Premium
- The Trade-Off
- How to Choose
Flex Consumption
Flex Consumption (FC1) is the default plan. It scales to zero when idle, so you pay per execution and the cost is near zero when no work is running. On the Flex plan each Function App gets its own plan, and the runner code is delivered to each app's deployment container in your storage account at install time. The agent runner always runs on its own Flex Consumption plan, regardless of which plan you choose for the three script runners.
Premium
Premium (Elastic Premium, EP1 by default) keeps the runner always warm. The three script runners share a single Premium plan, and each app boots its code from the runner package store. Premium is the least-cost tier that supports both VNet integration and the 30-minute step timeout, so it suits workloads that need consistent, low-latency execution. You can choose a larger SKU (EP2 or EP3) for heavier workloads.
The Trade-Off
The choice is between cost and cold-start latency:
- Flex Consumption — lowest cost when idle, because it scales to zero, but a run that arrives after the runner has gone idle waits while the app wakes (cold-start latency)
- Premium — higher baseline cost because the plan stays warm, but no cold-start delay and steady throughput
How to Choose
Use Flex Consumption when runs are intermittent and a few seconds of wake-up latency is acceptable; it keeps cost proportional to actual use. Use Premium when runs are frequent or latency-sensitive and you want the runner ready at all times. The plan is set when you install the runner, and because the installer is idempotent you can change it later by re-running the deployment with the other plan. For the resources each plan provisions, see the article on what the runner installer deploys.
Comments
0 comments
Please sign in to leave a comment.