Module 14 cheat sheet
One-page revision summary for Advanced Banking Architecture: services, Azure mapping, CLI commands, limits, decision rules, exam traps and cost traps.
Assumes: AWS CLI v2 · .NET 10 LTS · Region us-east-1 · facts as stated in the Module 14 lessons (checked 2026-11-18). Verified against official sources on 2026-11-18.
A printable summary of Module 14. Figures marked in the lessons with a verify note can change: confirm them in the current AWS documentation before you rely on them in a design.
Key services
| Service | What it is for in a banking design |
|---|---|
| API Gateway + CloudFront + WAF + Shield | One controlled, protected entry per channel |
| Amazon Cognito | Customer sign-in, short tokens, step-up for risky actions |
| Step Functions Standard | Long, audit-visible flows: onboarding, payments, approvals |
| Step Functions Express | Short, high-volume flows (up to 5 minutes) |
| DynamoDB | Idempotency keys, balance reservations, projections |
| Aurora | Double-entry ledger |
| Kinesis Data Streams + Lambda | Ordered per-account event streams, fraud rules |
| EventBridge + SQS | Domain events (a bus per domain) and a queue per consumer with a dead-letter queue |
| S3 + Glue + Athena | Reports, reconciliation files, log archive queries |
| S3 Object Lock | Tamper-resistant evidence in a separate log archive account |
| Direct Connect, Site-to-Site VPN, Amazon MQ | Hybrid link to core banking and MQ-style messaging |
| AWS Backup, Aurora Global Database | DR building blocks, immutable cyber recovery copies |
.NET / Azure → AWS
| Microsoft / Azure | AWS |
|---|---|
| API Management + App Service | API Gateway + ECS Fargate or Lambda (VPC Link) |
| Service Bus + Durable Functions | EventBridge + SQS + Step Functions |
| Service Bus sessions, duplicate detection | SQS FIFO groups and deduplication (5-minute window) |
| Azure SQL ledger tables | Aurora double-entry design, DynamoDB idempotency |
| Stream Analytics | Kinesis Data Streams + Lambda, Managed Service for Apache Flink |
| Storage + Data Factory reporting | S3 + Glue + Athena |
| Front Door + WAF | CloudFront + AWS WAF + Shield |
| Azure Site Recovery / paired Regions | Backup, Aurora Global Database, Route 53, your own Region pairing |
Most useful CLI commands
| Task | Command |
|---|---|
| Run a state machine and wait | aws stepfunctions start-sync-execution --state-machine-arn <arn> --input file://in.json |
| List state machines | aws stepfunctions list-state-machines --profile learn-sandbox --region us-east-1 |
| Read an idempotency item | aws dynamodb get-item --table-name hl-m14-idempotency --key file://key.json |
| Stream health (iterator age) | aws cloudwatch get-metric-statistics --namespace AWS/Kinesis --metric-name GetRecords.IteratorAgeMilliseconds … |
| Query logs with SQL | Athena over CloudTrail in S3 (not CloudTrail Lake) |
| Anything left tagged after a lab? | aws resourcegroupstaggingapi get-resources --tag-filters Key=Project,Values=cea-lab |
| Tear down the lab stack | cdk destroy from labs/m14-payment-saga-failures/infra/cdk |
Key limits and defaults
| Item | Value | Lesson |
|---|---|---|
| Step Functions Standard max time | 1 year; history kept 90 days after close | API orchestration |
| Step Functions Express max time | 5 minutes | API orchestration |
| State input or output | 256 KiB; pass references | API orchestration |
| Standard execution history | 25,000 events | API orchestration |
| REST API sync timeout (default) | About 29 s ⚠ VERIFY | Payment processing |
| Kinesis per shard | About 1 MB/s or 1,000 records/s in, 2 MB/s out ⚠ VERIFY | Real-time |
| SQS FIFO dedup window | 5 minutes | Payment processing |
| DynamoDB TTL deletion | Can lag by hours; do not rely on exact expiry | Payment processing |
Decision rules
| If… | Then… |
|---|---|
| Flow outlives a request, needs approval or audit history | Step Functions Standard |
| Flow is short, pure and unit-testable | In-service code |
| Scheme or core times out | Outcome is unknown: query by business reference before retry or refund |
| Before the scheme accepts / after | Compensate (release the hold) / move forward only (settle, later return) |
| Order matters | Per account: partition key or FIFO group, never global |
| Any consumer, any queue | Assume at-least-once, make it idempotent |
| Model call in a fraud path | Short timeout plus a monitored fallback rule set |
| Audit logs | Separate account, Object Lock, validation, protected key |
| Choosing a DR strategy | Start from impact tolerance, pick the cheapest that meets measured RTO and RPO |
Common exam traps
| Trap | Reality |
|---|---|
| Amazon Fraud Detector for new fraud design | Closed to new customers (2025-11-07): rules on Kinesis and Lambda, SageMaker AI or Bedrock |
| CloudTrail Lake or Audit Manager for new designs | Closed to new customers (2026-05-31) / no new onboarding (2026-04-30): CloudTrail to S3 with Athena, own evidence catalogue |
| Express workflows for long human approvals | Express stops at 5 minutes: use Standard |
| Exactly-once from EventBridge or SQS standard | At-least-once, unordered: idempotent consumers |
| Retry after a timeout | Can double-pay: enquire first, use idempotency keys |
| "Compliant by using AWS" | Never: design shows touchpoints, specialists sign off |
Cost traps
| Trap | Why |
|---|---|
| Step Functions Standard at volume | Billed per state transition |
| Kinesis shards left running | Hourly billing: stop within the hour or simulate locally |
| Direct Connect, Amazon MQ, multi-Region DR | Run all day, design-only in a sandbox |
| SageMaker endpoints for scoring | Always-on instances: design-only or a local rules engine |
| Verbose logging of payloads | Storage and personal-data risk |
| Lab stack not destroyed | Run cdk destroy, then check tagged resources |
Independent learning resource — not affiliated with, endorsed or sponsored by Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates. All banking data in examples is fictional. Sample architectures are for learning and are not automatically compliant with GDPR, PCI DSS or any banking regulation.
Disaster-recovery architecture for important business services
Map important business services and their impact tolerances to backup-and-restore, pilot-light, warm-standby or active-active designs, plan for cyber recovery and dependencies, and check the mapping with a runnable C# tolerance checker.
Module 15: Multi-Account Governance & FinOps
Set up and govern many AWS accounts the way a bank must: organisational units, guardrails, central logging and security tooling, and evidence an auditor can use. You also learn FinOps, so every team can see, explain and reduce what it spends.