Fleet-Level Analysis — Thinking at Scale

Why Per-Object Math Isn't Enough

The break-even calculation tells you whether S3-INT is profitable for a single object. But in practice, you're making a bucket-level decision that applies to millions of objects with a distribution of sizes. A bucket might have 80% of its objects below 250 KB and 20% above — and that 20% might hold 95% of the total storage volume.

The Right Mental Model: Weighted Average

The question to ask is not "what is the average object size?" but rather "what percentage of my storage volume (not object count) is in objects above 250 KB?"

Consider two buckets, both containing 10 TB of data:

Bucket A — Log Archive:
- 50 million objects, average size 200 KB
- 90% of objects are below 250 KB
- Those small objects represent 60% of total volume
- Only 40% of volume is in large objects → modest savings

Bucket B — Media Archive:
- 500,000 objects, average size 20 MB
- 100% of objects are above 250 KB
- 100% of volume is eligible for tiering → strong savings

Both buckets are 10 TB. But Bucket B is a far better candidate for S3-INT.

The Fleet-Level Formula

To estimate net monthly savings for a bucket:

Monitoring Cost = (Total Object Count / 1,000) × $0.0025

Eligible Volume (GB) = Volume in objects > 250 KB
Storage Savings = Eligible Volume × $0.0105

Net Monthly Impact = Storage Savings - Monitoring Cost

If Net Monthly Impact is positive, S3-INT is worth enabling. If negative, use Lifecycle Policies instead.

Pulling the Data You Need

You can get object count and size distribution from S3 Storage Lens (AWS Console → S3 → Storage Lens) or by running an S3 Inventory report. The key metrics are:

• StorageBytes — total volume
• ObjectCount — total object count
• Size distribution histogram (available in Storage Lens dashboards)