S3 is the simplest money most teams are leaving on the table in AWS
Not because S3 is expensive. Because almost everyone uses it the same way regardless of what they're storing and how often they access it
Everything goes into Standard storage class. The logs from two years ago. The backups from before the last major architecture change. The ML training datasets that were used for one experiment and haven't been touched since. The user uploads from accounts that were deleted. All of it. Standard. Full price. Every month
S3 has six storage classes for a reason
Standard is for data you access frequently. Standard-IA is for data you access occasionally - about 40 percent cheaper. Glacier Instant Retrieval is for archives you might need within milliseconds - about 70 percent cheaper than Standard. Glacier Deep Archive is for data you almost never need - about 95 percent cheaper than Standard
The difference in cost between storing a terabyte in Standard versus Glacier Deep Archive is significant at any scale
The fix is lifecycle policies. You define rules: after 90 days without access, move to Standard-IA. After 180 days, move to Glacier. After a year, move to Deep Archive or delete. AWS handles the transitions automatically
The caveat: retrieval costs money on the cheaper classes. If you're moving data to Glacier and then accessing it constantly, you've created a more expensive problem than the one you solved. Know your access patterns before you set the policies
The second thing to check: S3 Intelligent-Tiering
For data where access patterns are unpredictable, Intelligent-Tiering automatically moves objects between access tiers based on actual usage. No retrieval fees. No access pattern analysis required. You pay a small monitoring fee per object and AWS optimizes the rest
In most accounts with significant S3 spend, lifecycle policies and Intelligent-Tiering together reduce storage costs 30 to 50 percent
The data is the same. The bill is not
Top comments (0)