Why AWS Added This Rule
When S3 Intelligent-Tiering launched, customers discovered that enabling it on buckets full of tiny files resulted in monitoring fees with zero savings — because the storage savings on a 5 KB file are fractions of a cent per year. AWS responded by adding an automatic exemption: objects smaller than 128 KB are excluded from the tiering system entirely.
What the < 128 KB Rule Actually Does
| Behavior | Objects < 128 KB | Objects ≥ 128 KB |
|---|---|---|
| Monitoring fee charged | ❌ No | ✅ Yes |
| Eligible for IA tier | ❌ No | ✅ Yes |
| Billed at | Standard rates | Tier-appropriate rates |
| Affected by access patterns | ❌ No | ✅ Yes |
Objects below 128 KB sit permanently in the Frequent Access tier, billed at standard S3 rates. They are invisible to the tiering engine.
The Implication: S3-INT Is Useless for Small-Object Buckets
The 128 KB rule protects you from paying monitoring fees on tiny files, but it also means Intelligent-Tiering provides zero benefit for buckets dominated by small objects. Enabling S3-INT on a bucket of 100 million 50 KB log files will:
- Cost you $0 in monitoring fees (objects are exempt)
- Save you $0 in storage costs (objects never move tiers)
- Accomplish nothing
The correct tool for small-object buckets is a Lifecycle Policy — a rule that moves or expires objects based on age, with no per-object monitoring fee. For example, a policy that transitions all objects to S3 Glacier after 30 days applies to any object regardless of size, and costs nothing to evaluate.
A common interview scenario: "You enabled S3 Intelligent-Tiering on a bucket six months ago and your bill hasn't changed. Why?" The answer has two possible causes: (1) most objects are below 128 KB and are exempt from tiering, or (2) objects are being accessed frequently enough that they never qualify for the IA tier. Both are diagnosable with S3 Storage Lens metrics. The follow-up question is usually: "What would you do instead?" — answer with Lifecycle Policies for small objects and access pattern analysis for frequently-accessed data.