S3 Intelligent-Tiering
Key concepts
Automatic tier transitions
Access tier types
Monitoring fee
Archive access tiers
When to use vs lifecycle
Overview
Amazon S3 Intelligent-Tiering is the only cloud storage class that automatically moves data between access tiers based on changing access patterns, delivering cost savings without performance impact, retrieval fees, or operational overhead.
Core Concept
S3 Intelligent-Tiering automatically moves objects between tiers based on access patterns. Objects not accessed for 30 days move to Infrequent Access (40% savings), and after 90 days to Archive Instant Access (68% savings). When accessed, objects automatically return to Frequent Access with no retrieval fees.
For exam questions about 'unknown' or 'unpredictable' access patterns, S3 Intelligent-Tiering is almost always the correct answer. Remember: no retrieval fees for automatic tiers, only the optional Archive/Deep Archive tiers have retrieval costs.
Key Concepts
Access Tier Architecture
Automatic Access Tiers
Frequent Access Tier (Default)
- All objects start here
- Same performance as S3 Standard
- Same price as S3 Standard (~$0.023/GB)
Infrequent Access Tier (30 days no access)
- Automatic transition after 30 consecutive days of no access
- ~40% lower storage cost than Frequent Access
- Same millisecond access latency
- No retrieval fees
Archive Instant Access Tier (90 days no access)
- Automatic transition after 90 consecutive days of no access
- ~68% lower storage cost than Frequent Access
- Same millisecond access latency
- No retrieval fees
Key Behavior: Objects automatically move back to Frequent Access when accessed - no manual intervention required.
Optional Archive Tiers
Archive Access Tier (Optional)
- Must be explicitly enabled
- Configurable: 90-730 days before transition
- Asynchronous access (3-5 hours retrieval)
- Retrieval fees apply (similar to Glacier Flexible)
- Best for data that can tolerate retrieval delay
Deep Archive Access Tier (Optional)
- Must be explicitly enabled
- Configurable: 180-730 days before transition
- Asynchronous access (12-48 hours retrieval)
- Lowest storage cost
- Retrieval fees apply (similar to Glacier Deep Archive)
Important: Optional tiers require bucket-level or prefix-level configuration to enable.
Monitoring and Automation Fee
Cost: $0.0025 per 1,000 objects per month (~$2.50 per million objects)
What It Covers:
- Access pattern monitoring
- Automatic tier transitions
- No additional transition charges between tiers
Object Size Consideration:
- Objects < 128 KB are NOT monitored
- Small objects always stay in Frequent Access tier
- Small objects are NOT charged the monitoring fee
- For buckets with many small files, consider Standard instead
Calculation Example:
- 10 million objects = $25/month monitoring fee
- If 50% move to IA tier: ~$115/month savings on 10 TB
- Net benefit: $90/month savings
Tier Transition Timeline
S3 Intelligent-Tiering Access Tiers
| Tier | Days No Access | Storage Savings | Retrieval Fee | Access Latency | Configuration |
|---|---|---|---|---|---|
| Frequent Access | 0 (default) | 0% | None | Milliseconds | Automatic |
| Infrequent Access | 30 days | ~40% | None | Milliseconds | Automatic |
| Archive Instant Access | 90 days | ~68% | None | Milliseconds | Automatic |
| Archive Access | 90-730 days | ~71% | Yes | 3-5 hours | Optional (enable) |
| Deep Archive Access | 180-730 days | ~95% | Yes | 12-48 hours | Optional (enable) |
Intelligent-Tiering vs Lifecycle Policies
Intelligent-Tiering vs Lifecycle Policies
| Factor | S3 Intelligent-Tiering | Lifecycle Policies |
|---|---|---|
| Best For | Unknown/changing access patterns | Predictable access patterns |
| Management | Fully automatic | Manual rule configuration |
| Monitoring Fee | $0.0025/1,000 objects/month | None |
| Transition Charges | None between tiers | $0.01/1,000 objects per transition |
| Retrieval Fees | None (automatic tiers) | Depends on destination class |
| Object Deletion | Not supported | Supported (expiration actions) |
| Min Object Size | 128 KB for auto-tiering | No restriction (but 128 KB billing in IA) |
| Move Back Up | Automatic on access | Not possible with lifecycle |
{
"Id": "ArchiveTierConfiguration",
"Status": "Enabled",
"Tierings": [
{
"Days": 90,
"AccessTier": "ARCHIVE_ACCESS"
},
{
"Days": 180,
"AccessTier": "DEEP_ARCHIVE_ACCESS"
}
]
}# Create Intelligent-Tiering configuration with archive tiers
aws s3api put-bucket-intelligent-tiering-configuration \
--bucket my-bucket \
--id "MyConfig" \
--intelligent-tiering-configuration '{
"Id": "MyConfig",
"Status": "Enabled",
"Tierings": [
{"Days": 90, "AccessTier": "ARCHIVE_ACCESS"},
{"Days": 180, "AccessTier": "DEEP_ARCHIVE_ACCESS"}
]
}'
# List Intelligent-Tiering configurations
aws s3api list-bucket-intelligent-tiering-configurations \
--bucket my-bucket
# Get specific configuration
aws s3api get-bucket-intelligent-tiering-configuration \
--bucket my-bucket \
--id "MyConfig"
# Delete configuration
aws s3api delete-bucket-intelligent-tiering-configuration \
--bucket my-bucket \
--id "MyConfig"# Upload new object directly to Intelligent-Tiering
aws s3 cp myfile.txt s3://my-bucket/myfile.txt \
--storage-class INTELLIGENT_TIERING
# Copy existing object to Intelligent-Tiering
aws s3 cp s3://my-bucket/existing.txt s3://my-bucket/existing.txt \
--storage-class INTELLIGENT_TIERING
# Use lifecycle rule to transition existing objects
# (see lifecycle policies topic for details)Cost Analysis Scenarios
When Intelligent-Tiering Saves Money
Scenario 1: Mixed Access Patterns
- 1 TB of data, 50% accessed monthly, 50% rarely accessed
- Standard: $23.55/month
- Intelligent-Tiering: ~$14/month (40% of data in IA tier) + $2.50 monitoring = ~$16.50/month
- Savings: ~$7/month (30%)
Scenario 2: Unknown Patterns (Data Lake)
- 100 TB data lake, access patterns vary by team
- Risk with Standard-IA: Retrieval fees if access increases
- Intelligent-Tiering: Automatic optimization, no retrieval fees
- Benefit: Cost optimization without risk
Scenario 3: Seasonal Business Data
- Data accessed heavily quarterly, idle between
- Lifecycle can't handle dynamic patterns
- Intelligent-Tiering adjusts automatically
- Benefit: Matches real access patterns
When Intelligent-Tiering May NOT Be Best
Scenario 1: Known Access Patterns
- Logs accessed for 7 days, then never
- Lifecycle: Standard → Glacier (defined rule)
- IT: Pays monitoring fee for predictable data
- Better Choice: Lifecycle policies
Scenario 2: Many Small Files
- Millions of files < 128 KB each
- Small files not eligible for auto-tiering
- Pay Frequent Access price regardless
- Better Choice: S3 Standard (no monitoring fee)
Scenario 3: Always Hot Data
- Data accessed multiple times daily
- Will never transition to lower tiers
- Monitoring fee adds no value
- Better Choice: S3 Standard
Best Practices
- Use for Unknown Patterns: Default to Intelligent-Tiering when you can't predict access patterns
- Enable Archive Tiers: For long-term data, enable optional Archive/Deep Archive tiers for maximum savings
- Analyze Object Sizes: Avoid Intelligent-Tiering for buckets dominated by small files (< 128 KB)
- Combine with Lifecycle: Use lifecycle to transition objects TO Intelligent-Tiering, let IT handle optimization
- Monitor Costs: Review Cost Explorer to validate savings exceed monitoring fees
- Consider First-Month Costs: Objects start in Frequent Access; savings begin after 30 days
- Use Prefix Filters: Enable archive tiers only for specific prefixes if needed
Common Exam Scenarios
Exam Scenario Decision Guide
| Scenario | Recommended Solution | Key Reasoning |
|---|---|---|
| Media files with unpredictable access patterns | S3 Intelligent-Tiering | Automatic optimization, no retrieval fees when accessed |
| Data lake with multiple teams, varying access | S3 Intelligent-Tiering | Handles mixed patterns without manual rules |
| Log files: 7 days active, then archived forever | Lifecycle to Glacier | Predictable pattern, no monitoring fee needed |
| Backup data accessed only for DR testing | Intelligent-Tiering with Archive tiers | Rare access benefits from archive, instant access when needed |
| Millions of small config files | S3 Standard | Small files not auto-tiered, avoid monitoring fee |
| New application, access patterns unknown | S3 Intelligent-Tiering | Let AWS optimize while you learn patterns |
| Compliance data: 1 year retention, rarely accessed | Intelligent-Tiering with Deep Archive tier | Automatic archiving with compliance retention |
| Frequently accessed API cache data | S3 Standard | Always hot data gains nothing from IT |
Common Pitfalls
Small Objects Not Auto-Tiered
Objects smaller than 128 KB are always stored in the Frequent Access tier and are NOT monitored for tiering. They don't incur monitoring fees but also don't benefit from automatic cost optimization. For buckets with many small files, S3 Standard may be more cost-effective.
Optional Tiers Require Configuration
Archive Access and Deep Archive Access tiers are NOT enabled by default. You must explicitly create an Intelligent-Tiering configuration to activate them. Without configuration, objects only move between Frequent, Infrequent, and Archive Instant Access tiers.
First Month Higher Costs
When transitioning to Intelligent-Tiering, the first month will have higher costs than expected. Objects start in Frequent Access tier (same price as Standard) plus monitoring fees. Savings begin in month 2 when objects start moving to cheaper tiers.
Archive Tier Retrieval Fees
While the three automatic tiers (Frequent, Infrequent, Archive Instant) have NO retrieval fees, the optional Archive Access and Deep Archive Access tiers DO have retrieval fees similar to Glacier. Enable these only for truly archival data.
Cannot Delete Objects with IT
Intelligent-Tiering manages storage tiers but does NOT support automatic object deletion. You still need lifecycle expiration actions to delete objects after a retention period.
Related Services
Quick Reference
Tier Transition Summary
| From | To | Trigger | Retrieval Fee |
|---|---|---|---|
| Frequent Access | Infrequent Access | 30 days no access | None |
| Infrequent Access | Archive Instant Access | 90 days no access | None |
| Archive Instant Access | Archive Access | Configurable (90-730 days) | None (to move) |
| Archive Access | Deep Archive Access | Configurable (180-730 days) | None (to move) |
| Any Lower Tier | Frequent Access | Object accessed | None (automatic tiers) |
| Archive/Deep Archive | Frequent Access | Object accessed | Yes (restore fees) |
Key Pricing (US-East-1)
| Component | Cost |
|---|---|
| Frequent Access Storage | $0.023/GB/month |
| Infrequent Access Storage | $0.0125/GB/month |
| Archive Instant Access Storage | $0.004/GB/month |
| Monitoring Fee | $0.0025/1,000 objects/month |
| Transition to IT (from Standard) | $0.01/1,000 objects |
| Archive Access Retrieval | ~$0.03/GB |
| Deep Archive Retrieval | ~$0.02/GB |
CLI Quick Reference
# Put object in Intelligent-Tiering
aws s3 cp file.txt s3://bucket/file.txt --storage-class INTELLIGENT_TIERING
# Configure archive tiers
aws s3api put-bucket-intelligent-tiering-configuration \
--bucket BUCKET --id CONFIG_ID \
--intelligent-tiering-configuration file://config.json
# List configurations
aws s3api list-bucket-intelligent-tiering-configurations --bucket BUCKET
# Check object's current tier
aws s3api head-object --bucket BUCKET --key KEY \
--query '[StorageClass, ArchiveStatus]'
Test Your Knowledge
A company stores 50 TB of user-generated content in S3 Standard. Access patterns are completely unpredictable - some files are accessed daily, others not touched for months. Which storage strategy provides the BEST cost optimization with minimal operational overhead?
A data lake contains 100 million small JSON files (average 50 KB each) with varying access patterns. The team wants to optimize storage costs. What is the MOST cost-effective approach?
An organization enables S3 Intelligent-Tiering for a bucket and notices higher costs in the first month compared to S3 Standard. The access patterns haven't changed. What explains this?
A company wants to use S3 Intelligent-Tiering with Archive Access tier enabled for compliance data that must be retained for 7 years. Data is rarely accessed but must be retrievable within 4 hours when needed. Which configuration is correct?