S3 Intelligent-Tiering: When You Don't Know the Access Pattern, Use This

Key Takeaway

S3 Intelligent-Tiering is a storage class that automatically optimizes costs based on access patterns. Ideal for unpredictable or frequently changing access patterns, it can reduce costs by up to 68%.

Feature	S3 Intelligent-Tiering
Auto Tier Movement	Yes (based on access pattern)
Retrieval Fees	None
Minimum Storage Duration	None
Minimum Object Size	128KB (recommended)
Monitoring Cost	$0.0025/1,000 objects/month
Cost Savings	Up to 68%

---

What is S3 Intelligent-Tiering?

Concept

S3 Intelligent-Tiering is a storage class that monitors access patterns and automatically moves objects to the most cost-effective tier.

Intelligent-Tiering Core Value:
├── No access pattern analysis needed
├── No manual tier movement work
├── No performance degradation (instant retrieval)
├── No retrieval fees
└── Up to 68% cost savings

When to Use?

Intelligent-Tiering Suitable Cases:
✅ When access patterns are unpredictable
✅ When access patterns change frequently
✅ When Lifecycle policy management is burdensome
✅ Data lakes, new applications
✅ User-generated content (UGC)

Not Suitable Cases:
❌ When access patterns are clear (choose Standard or IA directly)
❌ Small objects under 128KB (monitoring cost inefficient)
❌ Short-term data deleted within 30 days

---

Access Tier Structure

Default Tiers (Auto-Enabled)

Default Access Tiers:
┌─────────────────────────────────────┐
│     Frequent Access (FA)            │
│     - Frequently accessed data      │
│     - Same price as Standard        │
└───────────────┬─────────────────────┘
                │ 30 days no access
                ↓
┌─────────────────────────────────────┐
│     Infrequent Access (IA)          │
│     - 40% cheaper                   │
│     - Immediately returns to FA on access │
└─────────────────────────────────────┘

Optional Archive Tiers

Optional Archive Tiers (Manual Activation):
┌─────────────────────────────────────┐
│     Archive Instant Access (AIA)    │
│     - Moves after 90 days no access │
│     - 68% cheaper                   │
│     - Millisecond retrieval (instant)│
└───────────────┬─────────────────────┘
                │ 90-730 days no access (configurable)
                ↓
┌─────────────────────────────────────┐
│     Archive Access (AA)             │
│     - 3-5 hour retrieval            │
│     - 71% cheaper                   │
└───────────────┬─────────────────────┘
                │ 180-730 days no access (configurable)
                ↓
┌─────────────────────────────────────┐
│     Deep Archive Access (DAA)       │
│     - Within 12 hour retrieval      │
│     - 95% cheaper                   │
└─────────────────────────────────────┘

---

Cost Structure

Storage Cost Comparison (us-east-1)

Tier	Storage Cost	Savings
Frequent Access	$0.023/GB	-
Infrequent Access	$0.0125/GB	46%
Archive Instant Access	$0.004/GB	83%
Archive Access	$0.0036/GB	84%
Deep Archive Access	$0.00099/GB	96%

Additional Costs

Cost Item	Price
Monitoring & Automation Cost	$0.0025/1,000 objects/month
Retrieval Cost (IA/AIA)	None
Archive Tier Restore Cost	Same as Glacier

Cost Calculation Example

1TB Data, Monthly Cost Comparison:

S3 Standard:
└── 1,000GB × $0.023 = $23/month

S3 Intelligent-Tiering (50% FA, 50% IA assumed):
├── FA: 500GB × $0.023 = $11.50
├── IA: 500GB × $0.0125 = $6.25
└── Monitoring: ~$2.50 (1 million objects assumed)
Total: $20.25/month (12% savings)

Intelligent-Tiering (20% FA, 80% AIA assumed):
├── FA: 200GB × $0.023 = $4.60
├── AIA: 800GB × $0.004 = $3.20
└── Monitoring: ~$2.50
Total: $10.30/month (55% savings)

---

Archive Tier Activation

Console Configuration

Archive Tier Activation Steps:
1. S3 Console → Select bucket
2. Properties tab → Intelligent-Tiering Archive configurations
3. Create configuration
   - Configuration name: config name
   - Filter (optional): filter by prefix/tag
   - Archive Access tier: set 90-730 days
   - Deep Archive Access tier: set 180-730 days
4. Save

AWS CLI Configuration

# Create archive tier configuration
aws s3api put-bucket-intelligent-tiering-configuration \
  --bucket my-bucket \
  --id archive-config \
  --intelligent-tiering-configuration '{
    "Id": "archive-config",
    "Status": "Enabled",
    "Tierings": [
      {
        "Days": 90,
        "AccessTier": "ARCHIVE_ACCESS"
      },
      {
        "Days": 180,
        "AccessTier": "DEEP_ARCHIVE_ACCESS"
      }
    ]
  }'

---

Intelligent-Tiering vs Lifecycle Policies

Comparison Table

Aspect	Intelligent-Tiering	Lifecycle Policy
Tier Movement Basis	Actual access pattern	Fixed duration
Auto Return	Yes (returns to FA on access)	No (manual or restore)
Management Burden	Low	High (policy design needed)
Cost Prediction	Difficult (varies with pattern)	Easy (fixed rules)
Suitable When	Pattern unclear	Pattern clear

Selection Criteria

Pattern clear + rare access → Lifecycle (Standard → IA → Glacier)
Pattern unclear + varies frequently → Intelligent-Tiering
Pattern clear + frequent access → Standard
Pattern clear + occasional access → Standard-IA (direct selection)

Using Together

Lifecycle + Intelligent-Tiering Combination:
├── On upload: Standard class
├── After 30 days: Transition to Intelligent-Tiering (Lifecycle)
└── After: Intelligent-Tiering manages automatically

Benefits of this approach:
├── Short-term data: Can be deleted quickly
├── Long-term data: Auto-optimized
└── Minimized monitoring costs

---

Object Size Considerations

Objects Under 128KB

Characteristics of objects under 128KB:
├── Always stays in Frequent Access tier
├── Never moves to Infrequent Access
├── Monitoring cost still charged
└── Conclusion: Inefficient

Recommendation:
├── Under 128KB: Use S3 Standard
└── 128KB and above: Use Intelligent-Tiering

Using Lifecycle Filter

// Only Intelligent-Tiering for objects 128KB and above
{
  "Rules": [
    {
      "ID": "move-large-objects-to-it",
      "Filter": {
        "ObjectSizeGreaterThan": 131072
      },
      "Status": "Enabled",
      "Transitions": [
        {
          "Days": 0,
          "StorageClass": "INTELLIGENT_TIERING"
        }
      ]
    }
  ]
}

---

Practical Use Scenarios

1. Data Lake

Data Lake Structure:
├── Raw Zone: Intelligent-Tiering
│   └── Ingested raw data, access pattern unclear
├── Processed Zone: Intelligent-Tiering
│   └── Processed data, analysis frequency varies
└── Curated Zone: Standard
    └── Frequently used final data

2. Media Archive

Media Content Management:
├── Recently uploaded: Frequent Access
├── 30 days no access: Infrequent Access (auto)
├── 90 days no access: Archive Instant Access (if configured)
└── When needed: Immediately streamable (no retrieval delay)

3. Log Storage

Log Data Optimization:
├── Recent logs: Frequent analysis → FA
├── 30+ day logs: Occasional analysis → IA
├── 90+ day logs: Compliance retention → AIA
└── When retrieval needed: Immediate Athena query possible

---

Monitoring and Analysis

S3 Storage Lens

Information Available with Storage Lens:
├── Data distribution by tier
├── Cost savings effect
├── Access pattern trends
└── Optimization recommendations

CloudWatch Metrics

Key CloudWatch Metrics:
├── BucketSizeBytes: Bucket size
├── NumberOfObjects: Object count
└── Filter by Storage Class available

---

SAA-C03 Exam Focus Points

Common Question Types

Scenario	Answer
"Access pattern unpredictable"	Intelligent-Tiering
"Access pattern changes frequently"	Intelligent-Tiering
"Auto cost optimization"	Intelligent-Tiering
"Cost savings without retrieval fees"	Intelligent-Tiering
"Reduce Lifecycle policy management burden"	Intelligent-Tiering
"Data lake default storage"	Intelligent-Tiering

Common Pitfalls

❌ Intelligent-Tiering has retrieval fees on access
   → No retrieval fees even from IA/AIA tiers

❌ All objects automatically move to archive tiers
   → Archive tiers require manual activation

❌ Objects under 128KB also move to IA
   → Under 128KB always stays in FA

❌ Intelligent-Tiering is always cheaper than Standard
   → Depends on monitoring cost + object count

❌ Intelligent-Tiering is optimal when pattern is clear
   → Direct class selection is more efficient when pattern is clear

---

Frequently Asked Questions (FAQ)

Q: Should I choose Intelligent-Tiering or S3 Standard-IA?

Choose Standard-IA if you know the access pattern, Intelligent-Tiering if you don't.

• Expected access about once every 30 days → Standard-IA (30-day minimum storage)
• Cannot predict access frequency → Intelligent-Tiering

Q: Isn't the monitoring cost burdensome?

$0.0025 per 1,000 objects per month. For large files, storage cost savings far exceed monitoring costs. However, monitoring costs can be burdensome if there are many small files.

Q: How do I retrieve objects from Archive Access tier?

Call the RestoreObject API or request restore from console. Restoration takes 3-5 hours (AA) or 12 hours (DAA). Archive Instant Access is instantly retrievable.

Q: Can I change existing Standard objects to Intelligent-Tiering?

Yes, you can change the storage class of existing objects using Lifecycle policies or S3 Batch Operations. PUT COPY request costs apply when changing.

Q: Can I set Intelligent-Tiering as the default storage class for a bucket?

Direct default setting is not possible, but you can specify the x-amz-storage-class: INTELLIGENT_TIERING header on upload, or transition immediately via Lifecycle policy.

---

Summary

S3 Intelligent-Tiering is the answer when "you don't know the access pattern":

1. Auto-tiering: Automatic movement based on access patterns
2. No retrieval fees: No additional costs even from IA/AIA
3. Instant access: Archive Instant Access also millisecond retrieval
4. 128KB+ recommended: Use Standard for small objects
5. Archive optional: AA/DAA require manual activation

Related Posts

• S3 Storage Classes Guide
• S3 Lifecycle Policies

References

• S3 Intelligent-Tiering
• Intelligent-Tiering Archive Configuration