Serverless architecture is a cloud computing model where developers build and run applications without managing underlying server infrastructure. In serverless computing, the cloud provider dynamically allocates compute resources on-demand, automatically scales based on traffic, and charges only for actual execution time and resources consumed. The model excels at event-driven workloads, enabling applications to react to triggers like HTTP requests, database changes, file uploads, scheduled tasks, and message queues. While "serverless" doesn't mean servers disappear — they're simply abstracted away — this paradigm shift allows teams to focus entirely on business logic rather than infrastructure management, making it particularly effective for APIs, data pipelines, real-time processing, and microservices where cost efficiency and automatic scalability are critical. In 2026, serverless has expanded into stateful workflows, AI inference pipelines, and edge computing, with platforms supporting durable long-running executions and WebAssembly runtimes with microsecond cold starts.
What This Cheat Sheet Covers
This topic spans 14 focused tables and 141 indexed concepts. Below is a complete table-by-table outline of this topic, spanning foundational concepts through advanced details.
Table 1: Core Serverless Concepts
Before touching any platform, it helps to fix the vocabulary that every serverless conversation circles back to. These are the building blocks — what a function actually is, why cold starts and warm starts behave so differently, how billing tracks execution rather than uptime, and the stateless contract that forces your data to live somewhere else.
| Concept | Example | Description |
|---|---|---|
AWS Lambda, Azure Functions | • Execution model where code runs in stateless containers triggered by events • cloud provider handles all infrastructure, scaling, and availability. | |
S3 upload → Lambda → DynamoDB | • Design pattern where functions execute in response to asynchronous events from various sources • enables loose coupling between components. | |
First invocation delay: 50ms–3s | • Latency incurred when a new execution environment is initialized • includes downloading code, starting runtime, and loading dependencies. | |
Subsequent invocations: <10ms | • Execution using an already-initialized environment • dramatically faster as runtime and dependencies are cached. | |
$0.20 per 1M requests + GB-s | • Billing based on actual execution time and resources consumed • no charges for idle time, making it cost-effective for variable workloads. |