Serverless and Function-as-a-Service Architecture Questions

Building applications on managed, event-triggered compute: functions-as-a-service (AWS Lambda, Azure Functions, Google Cloud Functions, Cloudflare Workers) and serverless containers. Covers the invocation lifecycle and cold starts (init vs handler work, provisioned concurrency, packaging, layers and container images), statelessness and externalizing state, execution limits (timeouts, memory, payload and /tmp size), concurrency and scaling behavior (account limits, burst scaling, protecting downstream databases with connection proxies and throttling), event sources and trigger semantics (at-least-once delivery, retries, idempotent handlers, dead-letter handling), composing functions with managed services and workflow orchestrators (Step Functions and equivalents), and serverless-specific observability, security (per-function IAM, secrets) and pay-per-invocation cost modeling. Includes when serverless fits versus containers or VMs, and vendor lock-in trade-offs. General compute selection, generic messaging patterns, and general idempotency theory are covered by their own topics.