Cloud Architecture Patterns — Quick Reference

Pattern → problem it solves (memorize this mapping, not just names)

Pattern	Solves

|---|---|

Multi-AZ Active-Active	A single AZ failing shouldn't take the service down
Hub-and-Spoke	N VPCs needing to talk becomes an unmanageable mesh as N grows
Strangler Fig	Can't safely rewrite a large legacy system in one release
CQRS	Read and write workloads need different scaling/consistency shapes
Saga	No single transaction can span multiple microservices' databases
Serverless Event-Driven	Fixed capacity wastes money or under-provisions for spiky load

If you need...	Consider

|---|---|

99.99%+ uptime, single region	Multi-AZ Active-Active
Many VPCs sharing services/on-prem connectivity	Hub-and-Spoke
To migrate a legacy system without a big-bang rewrite	Strangler Fig
High read volume, different shape than write model	CQRS
Multi-step process across independent microservices	Saga
Unpredictable/spiky load, don't want idle fixed capacity	Serverless Event-Driven
Regional-scale disaster tolerance	Multi-region (on top of Multi-AZ, not instead of it)

	Choreography	Orchestration

|---|---|---|

Coordinator	None — services react to events	Central coordinator (e.g. Step Functions)
Process visibility	Scattered across services	Visible in one place
Best for	Few steps, loose coupling	Many steps, complex compensation logic
New dependency introduced	None	Every step depends on the coordinator

Pattern	Purpose

|---|---|

Circuit breaker	Stop calling a repeatedly-failing dependency, fail fast
Bulkhead	Isolate resource pools per dependency
Graceful degradation	Pre-decided fallback behavior per feature, not improvised during an incident

Pattern	Real added cost

|---|---|

Multi-AZ Active-Active	~2x compute (both AZs at full capacity, not idle standby)
Hub-and-Spoke	Per-hour + per-GB Transit Gateway charges on top of connected resources
CQRS	Operating two data stores instead of one
Multi-region active-active	Cross-region data transfer + full duplicate infrastructure

Pillar	One-line question

|---|---|

Operational Excellence	Can you run and monitor this, with runbooks?
Security	Who can do what, is data encrypted?
Reliability	Can you recover from failure — have you tested that?
Performance Efficiency	Right resource for the job?
Cost Optimization	Paying only for what you use?
Sustainability	Minimum environmental impact for the workload?