Serverless

Serverless describes compute models where the provider runs and scales the servers and the customer pays per invocation or per request. Functions-as-a-service is the best-known form, but managed databases, object triggers, and edge functions follow the same pay-per-use, zero-capacity-planning philosophy.

The advantages are real: no idle cost, automatic scaling from zero to high traffic, and deployment units small enough to ship in minutes. The constraints matter equally — execution time limits, cold starts on infrequently used functions, and billing that becomes hard to predict at sustained high volume.

Serverless fits spiky, event-driven workloads and glue logic best. For steady, predictable traffic, reserved compute is usually cheaper per unit of work; run the numbers at your actual traffic profile before committing an entire architecture to per-invocation pricing.