Design Cost-Optimized Architectures
Cost-Optimized Storage and Data Movement
CoreOptimize storage model, class, lifecycle, backup, archive, size, and transfer path from access requirements.
Aligned to the current SAA-C03 exam guide, verified August 16, 2026.
Why this matters
Lowest storage cost is not the lowest capacity rate. A defensible design prices the required protocol, durability, availability, access and retrieval pattern, request behavior, retention, backup, growth, and transfer path, then removes waste without breaking any of them.
Must Know
- Choose object, block, or file from access semantics before comparing cost; an incompatible cheaper service is not cost-effective.
- S3 storage classes trade access frequency, resilience model, minimum storage/retrieval characteristics, and request/retrieval cost. Lifecycle rules automate known transitions and expiration.
- S3 Intelligent-Tiering fits changing or unknown access patterns when its monitoring/automation model is preferable to predicting transitions. Archive classes fit data whose retrieval time and cost meet the requirement.
- EBS cost depends on volume family, provisioned size, and performance settings. SSD fits random/transactional I/O; HDD fits eligible sequential throughput workloads.
- EFS and supported FSx families can use lifecycle/tiering or capacity modes appropriate to their service; protocol compatibility and active working set decide the real saving.
- AWS Backup centralizes plans and retention for supported resources; archive storage and backup are different because backup includes recoverable resource state and policy.
- DataSync fits online AWS storage movement, Transfer Family fits managed partner/client file-transfer endpoints, Storage Gateway fits ongoing hybrid access, and Snow Family fits offline movement when networks are impractical.
- Use Cost Explorer for spend trends, Budgets for threshold/forecast alerts, and Cost and Usage Report for detailed allocation/analysis. Tags and account structure make storage ownership visible.
- Requester Pays can shift supported S3 request and transfer charges to authenticated requesters; batching can reduce avoidable request overhead when the workload permits.
Compare and Distinguish
- S3 lifecycle vs Intelligent-Tiering: lifecycle wins when access decay is predictable; Intelligent-Tiering wins when access is unknown or changes and automatic tiering meets constraints.
- S3 Standard-IA/One Zone-IA vs archive classes: infrequent online access, lower-resilience single-AZ use, and archival retrieval are different requirements.
- Archive vs backup: archive stores infrequently accessed data; backup preserves recoverable resource state and versions according to recovery policy.
- EBS SSD vs HDD: random/latency-sensitive I/O versus eligible sequential throughput. Right-size both capacity and performance.
- EBS vs EFS vs FSx: attached block, elastic shared Linux file, or specific managed filesystem—protocol wins before price.
- DataSync vs Transfer Family vs Storage Gateway vs Snow: managed movement, external protocol endpoint, ongoing hybrid interface, or offline transfer.
Scenario examples
- Scenario: Access to S3 objects steadily declines after 30 and 90 days. Think: lifecycle transitions encode the predictable pattern.
- Scenario: Object access is unpredictable and operators cannot maintain rules. Think: Intelligent-Tiering may reduce manual guessing if its access and cost characteristics fit.
- Scenario: A large sequential log-processing volume is overprovisioned on SSD. Think: evaluate an eligible EBS HDD family and required throughput, not capacity alone.
- Scenario: A partner uploads daily by SFTP. Think: Transfer Family preserves the protocol; DataSync is not a partner-facing SFTP endpoint.
- Scenario: A dataset cannot finish online within the migration window. Think: Snow Family may be lower-risk and cost-effective after hard timing constraints eliminate the network.
Exam traps
- The coldest storage class is not cheapest if retrieval timing or fees violate the workload.
- One Zone-IA is not a substitute when multi-AZ resilience is required.
- Lifecycle transition is not a backup and replication is not retention history.
- A smaller EBS volume can lose required performance depending on configuration; rightsize both dimensions.
- Deleting backups to save cost can violate RPO, retention, or compliance.
- Requester Pays does not make anonymous requests free or remove all bucket-owner charges.
Key takeaways
- Honor protocol, resilience, retrieval, and retention before optimizing price.
- Match S3 class and lifecycle to observed access.
- Rightsize block/file capacity and performance together.
- Choose transfer method from volume, network, recurrence, and client protocol.
- Measure ownership and restore value before deleting data.
How it works
- S3 lifecycle evaluates object age and class; Intelligent-Tiering observes access and moves eligible objects among tiers according to service behavior.
- EBS provisions block capacity/performance; EFS and FSx expose managed file capacity with service-specific storage/tiering choices.
- AWS Backup creates protected recovery points under plans and retention.
- Transfer services move or expose data according to online, protocol, hybrid-cache, or offline models.
When to use it
- Use lifecycle for predictable age transitions, Intelligent-Tiering for uncertain access, and archive only when retrieval fits.
- Use AWS Backup for governed recoverability and S3 archive classes for object archival.
- Use DataSync for online movement, Transfer Family for external protocols, Gateway for recurring hybrid access, and Snow for constrained networks.
- Use CUR for detailed allocation, Cost Explorer for interactive trends, and Budgets for alerts.
Security and governance implications
- Encrypt data and backups and preserve key access for the retention period.
- Do not sacrifice required multi-AZ durability or immutability for storage savings.
- Use tags/account ownership and least privilege for lifecycle and deletion changes.
- Test restores before expiring older recovery points.
Operational and diagnostic signals
- For rising cost, separate capacity growth, request count, retrieval, replication, transfer, and orphaned resources.
- For lifecycle surprises, inspect object eligibility, age, version state, and rule overlap.
- For restores that miss RTO, inspect retrieval tier and full dependency restoration.
- For transfer overruns, compare actual throughput, retries, network path, and protocol overhead.
More detail
- Storage total cost includes capacity, requests, retrieval, early-deletion/minimum-duration characteristics where applicable, replication, transfer, backups, and operational effort.
- Lifecycle and retention should be based on data value and access evidence, not blanket deletion.
- Elastic capacity reduces pre-provisioning but does not eliminate access-tier, throughput, or stale-data optimization.
- Central billing and tags reveal cost but do not optimize architecture without ownership and action.
Ready for the quiz?
- When is Intelligent-Tiering better than lifecycle transitions?
- Why are archive and backup not interchangeable?
- What makes an EBS HDD family cost-effective?
- Which clue separates DataSync from Transfer Family and Storage Gateway?
- How do Cost Explorer, Budgets, and CUR differ?
Related objectives
- D4.1.K1
- D4.1.K2
- D4.1.K3
- D4.1.K4
- D4.1.K5
- D4.1.K6
- D4.1.K7
- D4.1.K8
- D4.1.K9
- D4.1.K10
- D4.1.K11
- D4.1.S1
- D4.1.S2
- D4.1.S3
- D4.1.S4
- D4.1.S5
- D4.1.S6
- D4.1.S7
- D4.1.S8
- D4.1.S9
- D4.1.S10