S3 Storage Classes
Inkwell's S3 bill has crept up. Most of it is years of old order invoices that nobody opens, stored at exactly the same price as the cover images shoppers load every second. S3 lets each object sit in a storage class priced for how it is actually used.
The rule
The colder the class, the cheaper to keep, the dearer to fetch.
Every class is designed for the same eleven-nines durability. What changes is price, access speed, and the fine print.
- S3 Standard — the default. Frequent access, millisecond reads, no retrieval fee, stored across at least three zones.
- S3 Intelligent-Tiering — for unknown or changing patterns. S3 moves objects between access tiers automatically for a small per-object monitoring fee.
- S3 Standard-IA (Infrequent Access) — cheaper storage, still millisecond reads, but a per-GB retrieval fee, a 30-day minimum storage charge, and small objects are billed as at least 128 KB.
- S3 One Zone-IA — like Standard-IA but kept in one zone, so cheaper; only for data you can recreate.
- S3 Glacier Instant Retrieval — archive pricing with millisecond reads, for data read about once a quarter. 90-day minimum.
- S3 Glacier Flexible Retrieval — retrieval takes minutes to hours. 90-day minimum.
- S3 Glacier Deep Archive — the cheapest storage of all; retrieval takes hours. 180-day minimum.
You pick a class per object when uploading, or let lifecycle rules move objects later (next lesson).
The shortcut
Ask two questions about any data: "How often is it read?" and "When it is read, how fast must it arrive?" Often → Standard. Rarely but instantly → an IA or Instant Retrieval class. Almost never, and hours are fine → Flexible Retrieval or Deep Archive. Unknown → Intelligent-Tiering.
Current cover images
Read every second. Standard.
The trap is the fine print. Putting busy data in Standard-IA can cost more than Standard once retrieval fees are added, and deleting an object from Deep Archive after one month still bills the full 180 days.
Drills
Answer fast — trust the rule.
Which class is the default for new objects?
Which class charges a per-GB retrieval fee but still reads in milliseconds?
Data is read almost never and can wait up to 12 hours when it is. Which class is cheapest to keep it in?
Why is One Zone-IA a poor choice for Inkwell's only copy of its invoices?
Inkwell's access pattern for a new data set is completely unknown. Which class fits?
Remember
- The colder the class, the cheaper to keep, the dearer to fetch.
- Two questions: how often is it read, and how fast must it arrive?
- Trap: retrieval fees and minimum storage periods can make the "cheap" class the expensive one.
