Choosing Between S3 Standard vs Glacier
🔍 Overview: S3 Standard vs Glacier storage classes
S3 Standard (and Standard‑IA)
- Purpose: Frequent or predictable access.
- Access latency: Milliseconds.
- Durability: 99.999999999% (“11 nines”) (AWS Documentation, cloudwithdj.com)
- Availability: ~99.99% (Standard), ~99.9% (Standard‑IA) (AWS Documentation)
- Cost: Higher per‑GB; minimal or no retrieval fees. Standard‑IA cheaper storage but charges per retrieval.
- Minimum durations: None for Standard; 30 days minimum for Standard‑IA/One‑Zone‑IA (AWS Documentation)
Glacier Storage Classes (within S3)
Three archival tiers, each offering the same 11‑nines durability as S3 Standard (AWS Documentation).
Tier | Min Storage Duration | Typical Access Frequency | Retrieval Time | Real‑time Access? |
---|---|---|---|---|
Glacier Instant Retrieval | 90 days | Quarterly or less | Milliseconds | ✓ (AWS Documentation) |
Glacier Flexible Retrieval (formerly Glacier) | 90 days | 1–2 times/year | Minutes to hours | ✗ (must restore) (AWS Documentation, Digital Cloud) |
Glacier Deep Archive | 180 days | ≤ once/year | 12–48 hours | ✗ (restore required) (AWS Documentation, Amazon Web Services, Inc.) |
- Storage cost: up to ~90% lower vs Standard (e.g. ~$0.004 /GB/mo) (Medium)
- Retrieval cost: charged per GB and request; faster tiers cost more (Wikipedia)
- Deletion charges: deleting objects before minimum duration still bills for full minimum term (AWS Documentation)
- Object size note: Glacier Instant requires ≥ 128 KB minimum (billed as 128 KB if smaller) (AWS Documentation)
✅ When to Use Which
Use S3 Standard or Standard‑IA when
- Data is accessed frequently or predictably (Standard: > monthly; IA: ~monthly).
- Immediate, high‑performance access is required.
- You want no additional retrieval latency or fees.
Use Glacier Instant Retrieval when
- Data is rarely accessed (e.g. quarterly or less), but must be available instantly if needed.
- Example: medical images, satellite imagery, media archives that might be served dynamically (AWS Documentation, Amazon Web Services, Inc.)
- Cost‑sensitive vs Standard‑IA for low access frequency.
Use Glacier Flexible Retrieval when
- Data access is very infrequent (once or twice a year).
- Retrieval delays of minutes to hours are acceptable (Expedited: 1–5 min; Standard: 3–5 h; Bulk: 5–12 h) (Amazon Web Services, Inc., Reddit, AWS Documentation)
- Use case: backups, disaster recovery, audit logs.
Use Glacier Deep Archive when
- Data is archived for long-term compliance (e.g. 7–10 years) or rarely needed (< once/year).
- Lowest possible storage cost outweighs retrieval latency (makes sense if you can wait 12‑48 hours) (Amazon Web Services, Inc., AWS for Engineers, cloudwithdj.com)
- Ideal replacement for tape libraries, regulatory archives, digital preservation.
⚙️ Lifecycle & Cost Considerations
- Automatically transition objects from Standard to Glacier via S3 Lifecycle policies based on age, tags, etc.
- Transition into Glacier has no minimum wait time, but billing still honors the 90/180‑day minimum (Medium, Reddit).
- Lifecycle transitions cost per request, especially when transitioning many small objects. Sometimes batching small files (e.g. zipping) helps (Reddit).
🧠 Quick Use‑Case Scenarios
- Website assets (images/videos): S3 Standard (low latency, frequent use).
- Monthly backup archives: Standard‑IA or Glacier Instant if access needs are rare but quick.
- Yearly audit logs or compliance data: Glacier Flexible Retrieval – cost effective with some delay.
- Multi-year retention archives (e.g. legal, financial): Glacier Deep Archive — ultra‑low cost, slow access is acceptable.
🧩 Summary Comparison
- Access frequency & latency: Standard (frequent/millisecond) → Instant Glacier (rare/millisecond) → Flexible (rare/minutes–hours) → Deep Archive (very rare/hours).
- Storage cost: Standard > Standard‑IA > Instant Retrieval > Flexible Retrieval > Deep Archive.
- Retrieval cost & delay: Standard has none; Glacier tiers trade cost for delay.
✔️ Recommendation Tips
- Define your data by access pattern, latency requirement, frequency of retrieval, and retention policy.
- Use S3 Lifecycle rules to automatically tier data over time.
- Monitor retrieval volumes and sizes to avoid surprising retrieval costs (especially for large restores).
- Batch small files to minimize request count and extra overhead.