AWS Exam Review

Design Cost-Optimized Architectures

Review SAA-C03 Domain 4 by connecting cost-optimized storage, compute, databases, networking, budgets, CUR, Savings Plans, lifecycle rules, NAT, and exam traps.

intermediate5 min readUpdated 2026-06-04CloudCertificationCostCapacityOperationsTradeoffs
Cost OptimizationStorage LifecycleCompute PurchasingDatabase CapacityNetwork Transfer CostCost ExplorerAWS BudgetsSavings Plans

After this, you will understand

Cost optimization is not penny-pinching; it is matching architecture spend to workload value, usage pattern, and reliability requirement.

Plain version

This domain asks whether you can choose storage, compute, database, and network designs that meet requirements at the lowest appropriate cost.

Decision pressure

Learners pick the cheapest visible option and ignore retrieval fees, data transfer, NAT processing, idle capacity, operational risk, or commitment lock-in.

Exam-ready model

Separate cost visibility, usage shape, right-sizing, lifecycle, commitment strategy, and transfer path before selecting the cheaper architecture.

Think before readingWhy can the cheapest storage class become expensive?
Retrieval fees, minimum storage duration, request charges, and access frequency can make an apparently cheap storage class cost more overall.

Reading in progress

This page is saved in your local study history so you can continue later.

Next: Cost-Optimized Architecture Trap Drills

Study path

Read these in order

Start with the mechanics, then move into the patterns that explain why the system is shaped this way.

  1. 1Design Secure ArchitecturesAWS Review
  2. 2AWS Cost And Usage Reportaws-services

Concepts Covered

  • SAA-C03 cost-optimized architecture domain
  • Storage lifecycle and access patterns
  • Compute purchasing options
  • Database capacity planning
  • Network transfer cost
  • NAT and endpoint cost traps
  • Cost Explorer and Budgets
  • CUR and cost allocation tags
  • Savings Plans and Reserved Instances
  • Cost versus resilience tradeoffs

1. Domain Mental Model

Cost optimization is architecture fit.

The core question is:

what is the lowest-cost design that still satisfies the real requirement?

The cheapest service choice is not always the cost-optimized design. S3 Glacier classes can be wrong when immediate access is required. One NAT gateway can be cheaper but less available. Spot Instances can be cheap but wrong for interruption-intolerant workloads. Reserved capacity can save money but becomes waste if demand disappears. A smaller database can be cheaper until it creates latency, retries, and outages.

SAA-C03 cost questions reward total-cost reasoning.

2. Official Task Map

AWS groups this domain into four task areas:

  • cost-optimized storage
  • cost-optimized compute
  • cost-optimized databases
  • cost-optimized network architectures

The official weighting is 20 percent of scored content.

Arcflow maps that to:

  • cost visibility and allocation
  • storage class and lifecycle selection
  • compute model and purchase strategy
  • database engine and capacity fit
  • data transfer and network path design
  • right-sizing and optimization recommendations

Cost optimization is the smallest domain by percentage, but it appears everywhere because every architecture decision has a bill.

3. What AWS Is Testing

AWS is testing whether you understand both service pricing shape and architecture tradeoffs.

For storage, expect S3 storage classes, lifecycle rules, archive retrieval, Intelligent-Tiering, EBS volume types, EFS/FSx cost, backups, snapshots, DataSync, Transfer Family, Storage Gateway, and migration transfer choices.

For compute, expect EC2 instance families, Spot Instances, Reserved Instances, Savings Plans, Auto Scaling, Lambda, Fargate, containers, hibernation, and production versus non-production availability needs.

For databases, expect RDS, Aurora, DynamoDB capacity modes, read replicas, serverless options, caching, retention, database migration, and engine fit.

For networking, expect NAT gateways, VPC endpoints, Direct Connect, VPN, Transit Gateway, VPC peering, CloudFront, Global Accelerator, DNS, cross-AZ traffic, cross-Region transfer, and throttling strategies.

4. Service And Concept Clusters

Start with visibility:

Then commitments and right-sizing:

Then storage and data lifecycle:

Then database and network costs:

5. Architecture Reasoning Patterns

Use a five-step cost trace:

visibility -> usage pattern -> right service -> right size -> right purchase model

Visibility comes first. Cost Explorer helps investigate; Budgets alerts; CUR or Data Exports support deep analysis; tags and accounts create attribution.

Usage pattern decides service fit. Frequent object access, infrequent access, archival retrieval, shared file access, relational writes, key-value reads, and streaming ingestion all have different cost shapes.

Right service matters before right size. Choosing S3 for object uploads avoids EC2 disk ownership. Choosing DynamoDB for key-value scale may reduce relational operational burden. Choosing Athena for occasional S3 queries avoids a running warehouse.

Right size follows metrics. Compute Optimizer, CloudWatch, database metrics, and load tests matter more than guesswork.

Purchase model comes last. Commit only after usage is stable enough.

6. High-Yield Comparisons

Cost Explorer vs Budgets vs CUR: analysis, alerting, and detailed export.

Savings Plans vs Reserved Instances vs Spot: commitment discount, reservation-style discount, and interruption-prone spare capacity.

S3 Standard vs Intelligent-Tiering vs IA vs Glacier classes: access frequency, retrieval requirements, minimum duration, and monitoring or retrieval costs.

NAT Gateway vs VPC endpoint: general outbound path versus private AWS service path that can reduce NAT processing cost.

Lambda vs Fargate vs EC2: request/event-driven billing, container service billing, and instance-hour control.

DynamoDB on-demand vs provisioned: unpredictable traffic flexibility versus predictable capacity cost control.

Read replicas vs bigger database: read scaling and locality versus vertical capacity.

Direct Connect vs VPN: dedicated private connectivity with setup/commitment versus lower-entry encrypted internet connectivity.

7. Scenario Triggers

"Analyze monthly cost by service or account" points to Cost Explorer.

"Alert when spend exceeds threshold" points to AWS Budgets.

"Detailed billing data for Athena analysis" points to CUR or Data Exports.

"Unknown S3 access pattern" points to Intelligent-Tiering.

"Archive with rare retrieval and flexible delay" points to Glacier classes.

"Private subnet high-volume S3 access through NAT" points to gateway endpoint cost reduction.

"Steady EC2 usage" can point to Savings Plans or Reserved Instances.

"Fault-tolerant flexible compute that can be interrupted" points to Spot.

"Unpredictable DynamoDB traffic" often points to on-demand mode.

"Reduce repeated database reads" can point to caching.

8. Common Traps

Do not use the cheapest storage class without checking retrieval time, retrieval cost, and minimum duration.

Do not buy commitments for unstable workloads.

Do not choose Spot for workloads that cannot tolerate interruption.

Do not send private S3 traffic through NAT when a gateway endpoint fits.

Do not treat Cost Explorer as a spending enforcement tool.

Do not choose CUR when the requirement is a simple visual investigation.

Do not optimize cost by removing required backups, logs, encryption, or redundancy.

Do not ignore cross-AZ and cross-Region data transfer.

Do not right-size from intuition. Use metrics and recommendations.

9. Study Path

Study in this order:

  1. AWS Cost Explorer
  2. AWS Budgets
  3. AWS Cost And Usage Report
  4. AWS Savings Plans
  5. AWS Compute Optimizer
  6. S3 Lifecycle And Storage Classes
  7. S3 vs EBS vs EFS vs Instance Store
  8. DynamoDB vs RDS vs Aurora
  9. NAT Gateway vs VPC Endpoints
  10. Multi-Account Cost Governance

Then revisit performance and resilience pages. Cost questions often ask for the cheapest design that still meets performance and availability constraints.

Review Design High-Performing Architectures, Design Resilient Architectures, Multi-Account Cost Governance, and Backup vs Replication Recovery Design.

Official AWS references:

What to study next

These links keep the session moving: read prerequisites first, then open the systems, concepts, and patterns that deepen this page.