AWS Backup Archive Tier

TL;DR: Most teams keep backups for years and restore almost none of them. Archive Tier moves aging AWS Backup recovery points onto Glacier-backed storage at roughly $1/TB-month — 5–10× cheaper than warm backup storage — via one lifecycle field on the backup plan: "warm for 90 days, then archive, delete after 7 years." The trade: restoring an archived backup is a two-step, 3–5 hour affair, which is fine for audits and terrible for 3 AM production incidents.

The numbers

• Warm backup storage: ~$0.05/GB-month · Archive: ~$0.001/GB-month
• 10 TB kept 7 years: $42,000 all-warm vs ~$840 with a 90-day warm window — a ~$41k swing per backup plan
• Field example: 50 TB across databases, 5-year retention: $2,500/month → $300/month ($132k saved over the window)
• Restore from archive: 3–5 h standard (minutes if you pay expedited), then a normal restore from warm — two steps

Do this

1. Open your longest-retention backup plan first (7-year HIPAA / 10-year SOX plans spend ~95% of their life cold — biggest win per edit).

2. Add the cold-storage transition to the rule: console → backup plan → rule → lifecycle → Transition to cold storage after N days; or in Terraform:

   rule {
     lifecycle {
       cold_storage_after = 90
       delete_after       = 2555   # must be ≥ cold_storage_after + 90
     }
   }
   

3. Pick the warm window from restore reality: "the oldest backup I might plausibly need within the hour" defines it. 30–60 days warm covers most rollback scenarios; DR-critical copies stay warm.

4. For Organizations: set this in org-level backup policies, not account by account — and enforce "any plan with retention > 1 year must archive" via AWS Config.

5. Compliance bonus: pair with Vault Lock (WORM) — immutable and cheap is the combination both the regulator and the CFO sign off on.

Gotchas

• 90-day minimum in archive: deleting early still bills the remainder; retention under ~120 days shouldn't archive at all. The console enforces cold + 90 ≤ expire.
• Not for S3 backups — S3 recovery points use S3's own lifecycle/Glacier classes instead. Archive Tier covers EBS, RDS/Aurora, DynamoDB, EFS, EC2 AMIs, VMware.
• Shortening retention later triggers the minimum-duration blip on already-archived points — expect a one-time charge on cleanup.
• Don't confuse it with S3 Glacier classes — same underlying economics, completely different console and lifecycle system.
• Watch the transition actually happen: Cost Explorer (service = AWS Backup, group by usage type) confirms recovery points are moving tiers.

Skip this if

• The backup is part of your fast-recovery DR path — a 3–5 hour retrieval during a regional incident is the wrong place to save $50.
• Retention is short (< ~120 days) — the archive minimum eats the benefit.
• You're on a third-party backup tool — Archive Tier is AWS Backup-specific; check that tool's own cold-tier story (or whether AWS Backup should own the cold copies). For the S3-object equivalent, see Glacier Deep Archive.