Serverless compute is a cloud execution model where infrastructure management is abstracted away, allowing developers to deploy code that runs in response to events with automatic scaling and pay-per-use billing. The three major providers—AWS Lambda, Azure Functions, and Google Cloud Functions—each offer distinct runtime support, pricing structures, and integration ecosystems, while sharing challenges like cold starts and execution time limits. Since late 2024 and into 2026, the landscape has shifted significantly: Lambda SnapStart now covers Python 3.12+ and .NET 8+ (reducing cold starts to sub-second), Lambda Durable Functions provides code-first multi-step workflows up to one year, Lambda Managed Instances extends the serverless model to EC2 hardware, and a key INIT phase billing change effective August 2025 means optimizing initialization is now directly tied to your compute bill. Understanding the nuances of concurrency models, deployment strategies, and observability patterns is critical: a function that cold-starts in 2 seconds might be fine for batch processing but unacceptable for user-facing APIs, and arm64 (Graviton) delivers 17–25% faster cold starts across every runtime at 20% lower cost than x86.
What This Cheat Sheet Covers
This topic spans 17 focused tables and 153 indexed concepts. Below is a complete table-by-table outline of this topic, spanning foundational concepts through advanced details.
Table 1: Execution Models
Serverless functions are not monolithic—they differ radically based on how they're invoked. Choosing the wrong invocation model leads to unexpected costs, silent failures, and timeout surprises; understanding each model's guarantees, retry behavior, and payload limits is the first step toward reliable serverless design.
| Model | Example | Description |
|---|---|---|
aws lambda invoke --function-name MyFunc--payload '{"key":"value"}' | • Caller waits for response • used for API Gateway, ALB, and direct SDK invocations • 6 MB request/response limit (200 MB with streaming) • timeout up to 15 minutes for Lambda, 60 minutes for Cloud Run. | |
aws lambda invoke --invocation-type Event--function-name MyFunc | • Lambda queues request and returns immediately • automatic retry up to 2 times • 1 MB max payload (increased from 256 KB in Q1 2026) • supports destination routing for success/failure. | |
curl https://<url-id>.lambda-url.us-east-1.on.aws/— direct HTTPS endpoint, no API Gateway | • Built-in HTTPS endpoint without API Gateway cost • supports IAM auth or public access (AuthType NONE) • supports response streaming and CORS • no additional charge beyond standard Lambda invocation pricing. | |
Lambda polls SQS queue every 1–20 seconds; processes batch of 1–10 messages | • Lambda pulls events from SQS, Kinesis, DynamoDB Streams, Kafka • configurable batch size and window • built-in partial failure handling • Provisioned Mode available for Kafka sources. |