Scalable and Loosely Coupled Architectures — SAA-C03 Study Guide

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing AWS Certified Solutions Architect – Associate premium course — no separate purchase.

Included in this chapter:

SQS Standard vs FIFO: throughput, ordering, deduplication windows
EventBridge rule syntax + content-based filtering patterns
Step Functions Standard vs Express: choosing by execution duration + cost
Lambda-SQS pairing: visibility timeout, batch size, partial-batch failure
Kinesis Data Streams: shard math, enhanced fan-out vs polling
Dead-letter queues: alarms, redrive policies, message inspection

Messaging primitives compared

Service	Pattern	Ordering	Retention	Consumers	Best for
SQS Standard	Queue	Best-effort	14 days max	1+ workers	Decoupled async work items, max throughput
SQS FIFO	Queue	Strict per MessageGroupId	14 days max	1+ workers (per group)	Order-critical work; 300 msg/s default (3 000 w/batching)
SNS	Pub/sub topic	N/A	Not retained (in-memory delivery)	Many subscribers	Fan-out to email, SMS, HTTP, Lambda, SQS
EventBridge	Event bus	N/A	Up to 24h archive (replayable)	Many rules + targets	Schema-routed events; SaaS integrations; cross-account
Kinesis Data Streams	Stream	Per shard	Up to 365 days	Many parallel consumers (KCL)	Ordered events with replay; real-time analytics
Step Functions	Workflow	Linear (or parallel branches)	Up to 1 year history (Standard)	Service integrations + Lambdas	Multi-step orchestration with retries

Decision tree

Cheat sheet

Cheat sheet

Pick the integration primitive by how consumers consume
Every queue needs a DLQ and retry policy or it loses messages
Set SQS visibility timeout to at least 6× the Lambda timeout
FIFO caps at 300/3 000 TPS unless high-throughput mode is on
SNS doesn't retain messages, so front it with SQS for durability
Use EventBridge archive and replay to rebuild or test consumers from history
Step Functions: Standard for durable orchestration, Express for high throughput
Use EventBridge Pipes to wire a source to a target without Lambda glue
Turn on SQS long polling to cut empty receives and cost
API Gateway throttles at the account, stage, and method levels
Scale an ASG by traffic with ALBRequestCountPerTarget target tracking
Use Step Functions Distributed Map for million-scale S3 parallel processing
Pause a workflow for an external callback with .waitForTaskToken
Combine Retry exponential backoff with Catch for fallback handling
Catch States.ALL on a Parallel state to intercept any branch failure
Use ReportBatchItemFailures so only failed SQS messages reprocess
DLQ retention must exceed the source queue's, or messages expire on arrival
Use EventBridge global endpoints for automatic multi-region failover
Attach a DLQ per EventBridge rule target to capture undeliverable events

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References