
Cargando...
From milliseconds to hours — choosing the right storage class saves money and passes exams
Access speed and cost are the two axes that define every storage decision
| Feature | S3 Instant access object storage at scale | S3 Glacier Lowest-cost archival with flexible retrieval |
|---|---|---|
Primary Use Case Exam scenarios hinge on retrieval frequency and tolerable latency — memorize these two axes | Active data: websites, apps, data lakes, analytics, backups needing fast restore | Long-term archival: compliance records, media archives, rarely accessed data |
Storage Classes Available All Glacier tiers are S3 storage classes — they live in S3 buckets, not separate Glacier vaults (unless using the legacy Glacier API) | S3 Standard, S3 Intelligent-Tiering, S3 Standard-IA, S3 One Zone-IA, S3 Express One Zone | S3 Glacier Instant Retrieval, S3 Glacier Flexible Retrieval, S3 Glacier Deep Archive |
Retrieval Latency Glacier Instant Retrieval is the trap — it has Glacier pricing but millisecond access. Deep Archive is the cheapest but slowest. | S3 Standard / Standard-IA / One Zone-IA: milliseconds. S3 Express One Zone: single-digit milliseconds | Instant Retrieval: milliseconds. Flexible Retrieval: 1–5 min (Expedited), 3–5 hrs (Standard), 5–12 hrs (Bulk). Deep Archive: 12 hrs (Standard), 48 hrs (Bulk) |
Minimum Storage Duration Early deletion fees apply if objects are deleted before the minimum duration — a common cost-optimization trap | Standard: none. Standard-IA: 30 days. One Zone-IA: 30 days. Intelligent-Tiering: 30 days | Glacier Instant Retrieval: 90 days. Glacier Flexible Retrieval: 90 days. Glacier Deep Archive: 180 days |
Minimum Object Size (billing) Storing millions of tiny files in Glacier Flexible Retrieval can cost MORE due to per-object overhead — always factor object size | Standard: no minimum. Standard-IA: 128 KB (smaller objects billed at 128 KB). One Zone-IA: 128 KB | Glacier Instant Retrieval: 128 KB. Glacier Flexible Retrieval: 40 KB (32 KB metadata + 8 KB index overhead). Deep Archive: 40 KB |
Retrieval Cost Bulk retrievals from Glacier Flexible Retrieval can be free — exam loves testing Expedited vs Bulk cost trade-offs | Standard: free GET. Standard-IA: per-GB retrieval fee. One Zone-IA: per-GB retrieval fee | Instant Retrieval: per-GB fee. Flexible Retrieval: Expedited (highest), Standard (moderate), Bulk (lowest/free). Deep Archive: per-GB fee for both tiers |
Durability One Zone-IA is the only S3 class without multi-AZ redundancy — it loses durability guarantee if AZ is destroyed | 99.999999999% (11 nines) across all classes except One Zone-IA | 99.999999999% (11 nines) for all Glacier classes |
Availability SLA Glacier Flexible Retrieval has HIGHER availability SLA than Glacier Instant Retrieval — counterintuitive but tested | Standard: 99.99%. Standard-IA: 99.9%. One Zone-IA: 99.5%. Express One Zone: 99.95% | Glacier Instant Retrieval: 99.9%. Glacier Flexible Retrieval: 99.99%. Deep Archive: 99.99% |
Bucket/Vault Model Modern Glacier usage = S3 storage classes in S3 buckets. Legacy Glacier API (vaults/archives) is a different service endpoint — exam distinguishes these | General purpose buckets (or directory buckets for Express One Zone). Bucket names: 3–63 chars, lowercase letters, numbers, hyphens, periods | Objects stored in S3 buckets via storage class setting. Legacy Glacier API uses 'Vaults' — separate construct, separate API |
Lifecycle Policy Integration Lifecycle transitions only go DOWN the cost ladder (to cheaper storage) — you cannot lifecycle FROM Glacier back to Standard automatically | Source of lifecycle transitions — objects start here and can be moved to cheaper tiers over time | Destination of lifecycle transitions. Objects can be transitioned: Standard → Standard-IA → Glacier Instant → Glacier Flexible → Deep Archive |
Direct Upload You CAN upload directly to Glacier storage classes — you do NOT have to go through Standard first. Many candidates get this wrong. | Direct PUT/multipart upload to any S3 class including Standard-IA and One Zone-IA | Can PUT directly to Glacier Instant Retrieval and Glacier Flexible Retrieval via S3 API. Deep Archive also supports direct PUT |
Restore Process Restoring from Glacier Flexible/Deep Archive creates a TEMPORARY copy in S3 Standard — original stays in Glacier. Restore duration is configurable. | No restore needed — GET returns object immediately | Instant Retrieval: no restore needed (milliseconds). Flexible Retrieval & Deep Archive: must initiate a Restore request; restored copy placed in Standard temporarily (1–365 days) |
Encryption Legacy Glacier vaults encrypt by default — no configuration needed. S3 buckets now also encrypt by default (SSE-S3) as of 2023 | SSE-S3 (default), SSE-KMS, SSE-C, DSSE-KMS. Client-side encryption also supported | Glacier storage classes inherit S3 encryption options. Legacy Glacier API encrypts all data at rest by default using AES-256 |
Access Control Vault Lock (legacy Glacier) enforces compliance controls that CANNOT be changed once locked — stronger than S3 Object Lock in some compliance scenarios | Bucket policies, IAM policies, ACLs (legacy), S3 Object Ownership, Block Public Access, VPC endpoints | Via S3: same as S3. Via legacy Glacier API: Vault Access Policies, Vault Lock (WORM), IAM policies |
WORM / Compliance Lock For SEC 17a-4, FINRA, CFTC compliance — both S3 Object Lock (Compliance mode) and Glacier Vault Lock are valid answers; scenario context matters | S3 Object Lock with Governance or Compliance mode. Retention periods and legal holds | Glacier Vault Lock (legacy API) — immutable policy once locked. S3 Object Lock available when using Glacier as S3 storage class |
Relative Storage Cost Deep Archive is the lowest-cost storage in AWS — exam tests this as the answer for 'cheapest storage for 7-year compliance archive' | Standard: highest. Standard-IA: ~40–50% less than Standard. One Zone-IA: ~20% less than Standard-IA | Instant Retrieval: ~68% less than Standard. Flexible Retrieval: ~83% less than Standard. Deep Archive: ~95% less than Standard (cheapest AWS storage) |
Query in Place / Analytics Glacier Select on Flexible Retrieval lets you query compressed/uncompressed CSV or JSON without full restore — cost and time saver on exams | S3 Select, S3 Inventory, S3 Storage Lens, S3 Analytics, Athena, Redshift Spectrum — all natively supported | S3 Select works on Glacier Instant Retrieval. Glacier Select available for Flexible Retrieval archives. Deep Archive does NOT support S3 Select |
Replication You can replicate FROM S3 Standard TO Glacier Flexible Retrieval in another region — useful for DR + cost optimization combo scenarios | Same-Region Replication (SRR) and Cross-Region Replication (CRR) supported for Standard, Standard-IA, One Zone-IA, Intelligent-Tiering | CRR/SRR can replicate TO Glacier storage classes. Glacier Flexible Retrieval and Deep Archive can be replication DESTINATIONS but not sources of active replication |
Event Notifications s3:ObjectRestore:Completed event is the correct trigger for post-restore processing workflows — tested in DVA-C02 and DOP-C02 | S3 Event Notifications to SNS, SQS, Lambda, EventBridge for PUT, DELETE, COPY, restore events | Glacier restore completion triggers S3 event notifications (s3:ObjectRestore:Completed) — key for automation |
Multipart Upload Max single PUT object size is 5 GB. Max object size overall is 5 TB (via multipart). These limits apply to all S3 storage classes including Glacier | Supported. Required for objects > 5 GB. Recommended for objects > 100 MB. Max object size: 5 TB | Supported via S3 API for all Glacier storage classes. Legacy Glacier API has its own multipart upload API |
Intelligent-Tiering Auto-Archive Intelligent-Tiering eliminates retrieval fees for frequent/infrequent tiers but charges a per-object monitoring fee — not cost-effective for objects < 128 KB | S3 Intelligent-Tiering can automatically move objects to Archive Access tier (≥90 days inactive) and Deep Archive Access tier (≥180 days inactive) | Intelligent-Tiering Archive tiers use Glacier Flexible Retrieval and Deep Archive infrastructure under the hood |
Summary
Use S3 Standard and its variants (Standard-IA, One Zone-IA, Intelligent-Tiering) when your data needs to be accessed within milliseconds to seconds and operational agility matters. Use S3 Glacier storage classes (Instant Retrieval, Flexible Retrieval, Deep Archive) when data is accessed rarely or never — archival compliance, long-term backups, and media vaults — where retrieval latency of minutes to hours is acceptable in exchange for dramatically lower storage costs. The decision tree is simple: if retrieval SLA is under 1 minute, stay in S3; if you can wait hours and want the cheapest storage on AWS, go Deep Archive.
🎯 Decision Tree
Need millisecond access always → S3 Standard | Need millisecond access but infrequent → S3 Standard-IA or Glacier Instant Retrieval | Access pattern unknown → S3 Intelligent-Tiering | Can wait 1–5 minutes, rarely accessed → Glacier Flexible Retrieval (Expedited) | Can wait 3–5 hours, compliance archive → Glacier Flexible Retrieval (Standard) | 7-year retention, almost never accessed, cheapest possible → Glacier Deep Archive | Single AZ acceptable, infrequent access, not critical → S3 One Zone-IA
Glacier Instant Retrieval is NOT slow — it retrieves in milliseconds just like S3 Standard-IA, but costs significantly less for storage. The exam uses 'Glacier' in the name to trick you into thinking retrieval is slow. Choose it when: data is accessed quarterly or less, millisecond retrieval is required, and you want Glacier-level storage pricing.
Deep Archive is the CHEAPEST storage in all of AWS (~95% cheaper than S3 Standard) with up to 48-hour Bulk retrieval. Any exam scenario mentioning '7-year retention', 'compliance archive', 'rarely if ever accessed', or 'lowest possible cost' should immediately signal Deep Archive as the answer.
Minimum storage duration charges are a cost-trap on exam scenarios: Standard-IA and One Zone-IA = 30 days, Glacier Instant Retrieval and Glacier Flexible Retrieval = 90 days, Glacier Deep Archive = 180 days. If a scenario asks about storing objects for 45 days cheaply, Glacier Flexible Retrieval (90-day minimum) would cost MORE than Standard-IA (30-day minimum) despite lower per-GB rate.
Glacier Flexible Retrieval has THREE retrieval tiers (Expedited: 1–5 min, Standard: 3–5 hrs, Bulk: 5–12 hrs). Expedited may require Provisioned Retrieval Capacity to guarantee availability during peak demand. Bulk is cheapest and sometimes free. The exam tests which tier to use given a cost vs. speed trade-off.
S3 Object Lock (Compliance mode) and Glacier Vault Lock both provide WORM compliance, but Vault Lock is for the legacy Glacier API (vaults/archives) and creates an immutable policy that cannot be changed after locking. For modern architectures using S3 storage classes, use S3 Object Lock. Exam scenarios mentioning 'vault' point to legacy Glacier API.
Restoring an object from Glacier Flexible Retrieval or Deep Archive does NOT move it — it creates a TEMPORARY copy in S3 Standard for 1–365 days. The original stays in Glacier. You pay Standard rates for the temporary copy. The s3:ObjectRestore:Completed event notification triggers when restore is done — use this for Lambda-based post-restore automation.
You CAN upload directly to any Glacier storage class via PUT — you do NOT need to upload to S3 Standard first and then transition. This is a common misconception. Direct PUT to Glacier Instant Retrieval, Flexible Retrieval, and Deep Archive is fully supported via the S3 API.
Glacier Flexible Retrieval has a HIGHER availability SLA (99.99%) than Glacier Instant Retrieval (99.9%). This is counterintuitive and appears in scenario-based questions about service reliability commitments.
The #1 exam trap is confusing 'Glacier' with 'slow retrieval only.' Glacier Instant Retrieval provides MILLISECOND access — identical to S3 Standard-IA in retrieval speed — but at a lower storage cost. Candidates eliminate it from consideration because of the word 'Glacier' and choose Standard-IA instead, losing points. Always evaluate all three Glacier tiers separately: Instant (ms), Flexible (min-hours), Deep Archive (hours).
CertAI Tutor · DEA-C01, SAP-C02, CLF-C02, SAA-C03, DVA-C02, DOP-C02, SCS-C02 · 2026-02-22
Services
Comparisons