LEARN. PRACTICE. GROW.Concept and mechanism
The compute target should match what the application actually requires. VMs provide operating-system control with additional responsibility. App Service can reduce server management for compatible web applications. Container Apps abstracts Kubernetes management but does not provide direct API access. When operators and CRDs are requirements, assess a compatible target such as AKS alongside skills and operations. Choosing containers does not automatically mean the team must administer a cluster.
Guided application
For large parallel task sets, assess Batch with suitable pools and scheduling. In any model, frontend scaling must respect dependency limits. If each worker uses two connections and only 24 are available, twelve workers exhaust that budget. Define maxima, signals, behavior under pressure, and who changes capacity. Include support, upgrades, and decommissioning costs alongside compute cost.
The vendor requires CRDs, but no Kubernetes team exists. The decision needs an operable platform or a validated requirement change.
Common pitfalls
Confusing managed with no operations; choosing by familiarity; increasing workers without assessing the database.
Related topics: Integration and change contracts · Networking and complete-service migration
Compatibility, operations, and capacity belong in the same decision.
Reference: Azure compute selection · AZ-305 objectives 2026-04-17