S3 lifecycle transition cost mistakes are request fees and minimum-duration charges that erase the storage-class savings you meant to buy. A lifecycle rule moves each eligible object into another class. Amazon S3 bills a lifecycle transition request per object, on top of whatever the destination class costs to store. The request does not care that the object is tiny. The storage savings do.
Rutagon reads that line before a rule is applied to a bucket of small objects. S3 cost optimization is the wider pass over storage, requests, and incomplete uploads. Storage class choice is which class you wanted. This article is the transition itself. The S3 pricing page is where the request row and the minimum-duration footnotes live. We use the row for the destination class in the Region on the bill.
Where S3 lifecycle transition cost mistakes show up
The usage report names the request lines. Requests-Tier4 counts lifecycle transitions into S3 Glacier Instant Retrieval, S3 Intelligent-Tiering, S3 Standard-IA, or S3 One Zone-IA. Requests-Tier3 counts lifecycle requests into S3 Glacier Flexible Retrieval or S3 Glacier Deep Archive, and it also counts standard S3 Glacier Flexible Retrieval restore requests. A spike in Tier3 is not automatically a lifecycle rule. Restores share that line. We split them before we delete a rule that was not the cause.
On the pricing page, US East (N. Virginia) lists lifecycle transition requests per 1,000. The rows we use are $0.01 into Standard-IA, One Zone-IA, and Intelligent-Tiering, $0.02 into Glacier Instant Retrieval, $0.03 into Glacier Flexible Retrieval, and $0.05 into Glacier Deep Archive. Another Region has its own row. One million transitions into Deep Archive at the N. Virginia row is $50 in requests, before any storage change. The same million into Standard-IA is $10 in requests. The object count is the multiplier. The byte count is not.
bill: Requests-Tier4 for IA, Intelligent-Tiering, Glacier Instant Retrieval, One Zone-IA
bill: Requests-Tier3 for Flexible Retrieval and Deep Archive transitions, mixed with standard Flexible Retrieval restores
n_virginia_per_1000: 0.01 IA and Intelligent-Tiering, 0.02 Glacier Instant Retrieval, 0.03 Flexible Retrieval, 0.05 Deep Archive
unit: one request per object transitioned
compare: request total against the storage delta for that class and Region Minimum storage duration is the second charge, and it is on the same pricing page. Standard-IA and One Zone-IA have a 30-day minimum. Objects deleted, overwritten, or transitioned out before 30 days pay the storage you used plus a prorated charge for the rest of the 30 days. Glacier Instant Retrieval and Glacier Flexible Retrieval have a 90-day minimum. Glacier Deep Archive has 180 days. A rule that lands objects in Standard-IA and another rule that moves them to Glacier days later still owes the remainder of the first minimum. The early move is not free because the second class is cheaper.
Small objects make the request row louder than the storage row. Standard-IA, One Zone-IA, and Glacier Instant Retrieval have a minimum billable object size of 128 KB. A smaller object can sit in the class and still be billed as 128 KB. Intelligent-Tiering can store objects under 128 KB, and those objects are not monitored. They stay on the Frequent Access rate, with no monitoring charge, so the transition request bought no tier change. Glacier Flexible Retrieval and Deep Archive add overhead on every archived object: 8 KB billed at Standard rates for the name and metadata you can list, and 32 KB billed at the destination rate for index data. A bucket of tiny objects pays that overhead per object, plus the transition request per object.
The lifecycle default skips objects under 128 KB so that request does not fire. A rule that sets ObjectSizeGreaterThan below that size opts the small objects back in. Each one is a transition request. Configurations saved before the default changed can still transition small objects into Flexible Retrieval and Deep Archive until someone edits the configuration. Editing the configuration is what picks up the skip. We check the rule that is actually on the bucket, not the rule someone remembers writing.
The check before a rule ships
We count objects and bytes in the prefix the rule will touch, by size band, before the rule is enabled. The bands that matter are under 128 KB, 128 KB to a few megabytes, and the large objects that can actually pay back a transition. The request estimate is the eligible object count times the per-1,000 rate for the destination class in that Region. The storage estimate uses the destination rate, the minimum billable size where the class has one, and the 40 KB of overhead where Flexible Retrieval or Deep Archive adds it. If the request estimate is larger than a month of storage delta, the rule waits.
Intelligent-Tiering savings are about access tiers after an object is already there. The transition into that class is still a Tier4 request at the $0.01 per 1,000 row in N. Virginia. We do not treat "move everything to Intelligent-Tiering tonight" as free. Objects that will never be large enough to leave Frequent Access should stay on Standard, where there is no transition request and no 128 KB minimum billable size.
Aggregation is the fix when the data truly must be archived and the objects are small. Many small objects become fewer large objects, and the transition request is paid per resulting object. The overhead of 8 KB plus 32 KB is also paid per resulting object. A lifecycle rule pointed at the original prefix will not do that aggregation. It will transition whatever is eligible, one request each. The job that packs the objects is the engineering change. The lifecycle rule runs after the pack, with a size filter that keeps the leftovers out.
We also put the minimum duration into the rule calendar. A transition into Glacier Instant Retrieval is not followed by a transition into Deep Archive inside 90 days unless someone has accepted the prorated Instant Retrieval charge on purpose. Two rules can be written that way. The bill will follow the rules. The pricing page is the source for the 30, 90, and 180 day minimums and for the N. Virginia request rows above. We do not copy a Standard-IA rate onto a Deep Archive transition.
Restore behavior stays visible because Tier3 mixes it with archival transitions. A lifecycle change that is supposed to cut Tier3, followed by a week of standard Flexible Retrieval restores, can show a flat line. We look at restore requests separately before calling the rule a failure. Deletes are free. A lifecycle expiration is not a transition request. Confusing expiration with transition sends people hunting for a request line that expiration does not create.
What we change, and what we leave
We leave a transition in place when the objects are large, the prefix is cold, and the request total is small next to the storage delta on the pricing page for that Region. Deep Archive at $0.05 per 1,000 is cheap on a million large objects and expensive on a million tiny ones. The same sentence with a different size distribution is a different decision. AWS cloud infrastructure work includes this comparison. It does not include a blended "archive savings" rate.
Rutagon runs it as a short bill review: Tier4 and Tier3, object counts by size, the destination class row, the minimum duration, and the 8 KB plus 32 KB overhead where it applies. The pricing page is the source for the dollars. The rule on the bucket is the source for what will actually transition.
Frequently Asked Questions
Which bill line is the lifecycle transition?
Requests-Tier4 is the transitions into Glacier Instant Retrieval, Intelligent-Tiering, Standard-IA, and One Zone-IA. Requests-Tier3 is the transitions into Glacier Flexible Retrieval and Glacier Deep Archive, and standard Flexible Retrieval restores sit on that same line. Split restores out before you treat Tier3 as a lifecycle problem.
Why did a cheaper storage class cost more?
The transition request is per object. In US East (N. Virginia) the pricing page lists $0.01 to $0.05 per 1,000 requests depending on the destination class. A large count of small objects pays that request, the minimum billable size where the class has one, and, for Flexible Retrieval and Deep Archive, 8 KB at Standard rates plus 32 KB at the destination rate. The storage delta has to beat that total.
What does the 128 KB filter change?
The default skip means objects under 128 KB are not transitioned, so the request is not billed for them. A smaller ObjectSizeGreaterThan opts them in. Configurations saved before the default changed can still send small objects to Flexible Retrieval and Deep Archive until the configuration is edited. We read the rule that is on the bucket.
Do we owe the old class if we transition again quickly?
Standard-IA and One Zone-IA have a 30-day minimum, Glacier Instant Retrieval and Flexible Retrieval have 90 days, and Deep Archive has 180 days. Leaving early adds a prorated charge for the remainder on top of the storage already used. A second rule that fires inside that window is how a "cheaper" class still invoices the first minimum.
Ask for an AWS bill review that prices the transition request against the storage delta before a lifecycle rule runs. Contact Rutagon at rutagon.com/contact, 907-841-8407, or contact@rutagon.com.