Compare Intelligent-Tiering against all-Standard, including the per-1,000-object monitoring charge and the 128 KB rule.
Intelligent-Tiering charges the tier rate for wherever each object currently sits, plus a monitoring fee that scales with object count rather than bytes. Objects smaller than 128 KB are exempt from monitoring but also never move out of Frequent Access, so a bucket of tiny files pays the Standard rate with no benefit. The rates above are user-supplied list prices to reconcile with the AWS S3 pricing page for your region. The break-even is a ratio of object count to bytes stored: at roughly 0.0025 USD per 1,000 objects a month, monitoring only pays for itself when the average object is large enough to actually be demoted to a cheaper tier.
S3 Intelligent Tiering
cost = Σ(tier share x GB x tier rate) + monitored objects ÷ 1000 x monitoring charge, where the shares are normalised to 100%.
cost = Σ(tier share x GB x tier rate) + monitored objects ÷ 1000 x monitoring charge, where the shares are normalised to 100%. Intelligent-Tiering charges the tier rate for wherever each object currently sits, plus a monitoring fee that scales with object count rather than bytes. Objects smaller than 128 KB are exempt from monitoring but also never move out of Frequent Access, so a bucket of tiny files pays the Standard rate with no benefit. The rates above are user-supplied list prices to reconcile with the AWS S3 pricing page for your region.
The break-even is a ratio of object count to bytes stored: at roughly 0.0025 USD per 1,000 objects a month, monitoring only pays for itself when the average object is large enough to actually be demoted to a cheaper tier.
This calculator takes 10 inputs: Data in Intelligent-Tiering, Objects in the bucket, Share of objects smaller than 128 KB, Share of data in Frequent Access, Share of data in Infrequent Access, Share of data in Archive Instant Access, Frequent Access price, Infrequent Access price, Archive Instant Access price, Monitoring and automation charge. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
When access patterns are already known, or when the bucket holds a huge number of very small objects. A plain lifecycle rule is cheaper for predictable data, and objects under 128 KB never tier down at all.