S3 storage class cost is more than the price per GB-month. Minimum storage duration, minimum billable object size, transition requests and retrieval fees decide whether a lifecycle rule saves money or adds cost. Whether you compare S3 storage classes or Azure Blob access tiers, model each data group by object size, lifetime and read pattern, not by the price table alone.

S3 storage classes compared

Minimum duration and size come from the S3 documentation; prices come from the AWS Price List API for eu-central-1 (Frankfurt) on 10 October 2026, in USD excluding VAT.

ClassMinimum durationMinimum billable sizeRetrievalStorage USD/GB-monthTransition per 1,000 objects
Standardnonenonefree0.0245 (first 50 TB)–
Intelligent-Tieringnonenone; no tiering below 128 KBfree except Expedited restore0.0245 / 0.0135 / 0.005 by tier, plus USD 0.0025 per 1,000 objects for monitoring0.01
Standard-IA30 days128 KBUSD 0.01/GB0.01350.01
One Zone-IA30 days128 KBUSD 0.01/GB0.01080.01
Glacier Instant Retrieval90 days128 KBUSD 0.03/GB0.0050.02
Glacier Flexible Retrieval90 days40 KB extra per objectExpedited 0.036, Standard 0.012, Bulk 0 USD/GB0.004050.036
Glacier Deep Archive180 days40 KB extra per objectStandard 0.024, Bulk 0.005 USD/GB0.00180.06

What the table compresses:

  • Intelligent-Tiering moves an object to Infrequent Access after 30 days without access and to Archive Instant Access after 90. It does not monitor objects under 128 KB; they stay in Frequent Access. The optional Archive Access and Deep Archive Access tiers need a restore before reading.
  • Glacier Flexible Retrieval and Deep Archive bill 40 KB of metadata per object: 32 KB at the Glacier rate and 8 KB at the Standard rate. A Flexible Retrieval restore takes 1 to 5 minutes (Expedited), 3 to 5 hours (Standard) or 5 to 12 hours (Bulk); Deep Archive takes up to 12 or 48 hours. While the restored copy exists, you pay for both the archived object and the temporary copy.
  • Early deletion, overwriting or transitioning before the minimum duration ends costs the prorated remainder.
  • Small objects under 128 KB are no longer transitioned by S3 Lifecycle by default since September 2024. Configurations created before then keep the old behaviour until you edit them; ObjectSizeGreaterThan and ObjectSizeLessThan let you set it yourself.

Azure Blob access tiers compared

Prices come from the Azure Retail Prices API for Germany West Central, LRS, in EUR excluding VAT, on 10 October 2026.

TierMinimum durationRetrievalStorage EUR/GB-monthWrites per 10,000Reads per 10,000
Hotnonefree0.0172 (first 50 TB)0.04750.0038
Cool30 daysEUR 0.0088/GB0.00880.0880.0088
Cold90 daysEUR 0.0264/GB0.0040.15840.088
Archive180 days, offlineEUR 0.0211/GB, high priority EUR 0.1144/GB0.00160.10564.7298, high priority 57.1982

Microsoft bills a move to a cooler tier as a write operation in the destination tier, and a move to a warmer tier as a read operation plus data retrieval in the source tier. Early deletion is prorated, and so is rewriting a whole blob inside the window. Rehydration from Archive at standard priority can take up to 15 hours for objects under 10 GB. Two limits are easy to miss: the archive tier, and with it the lifecycle action tierToArchive, is not supported in ZRS, GZRS or RA-GZRS accounts. And switching an account's default tier to a cooler one charges write operations for every blob whose tier is inferred.

Smart tier is generally available in public Azure regions with zone redundancy. It moves blobs to Cool after 30 days without access and to Cold after a further 60, charges a monitoring fee per object over 128 KiB, and charges no transition, retrieval or early deletion fees. It requires a GPv2 account with ZRS, GZRS or RA-GZRS, and it does not include Archive.

When a lifecycle rule pays off

Calculate per prefix or container, not per account. You need five inputs: volume V in GB, object count N, remaining lifetime L in months, data read per month R in GB, and the prices of source class A and target class B.

Transition cost  = N / 1000 × transition price into B
Monthly saving   = V × (price A − price B)
                 − extra GB for minimum size or metadata × their price
                 − R × retrieval price B
Pays off when      monthly saving × L > transition cost
                   and L exceeds the minimum duration of B

The formula shows why average object size matters so much: the transition price applies per object, the saving per GB. Many small objects rarely move at a profit; Microsoft's guidance is to pack small files into TAR or ZIP archives before tiering. If R is unknown or volatile, Intelligent-Tiering or smart tier can be a better fit than a fixed rule. There, moves between tiers and retrievals cost nothing, but a monthly monitoring fee applies per object, so model it with N rather than V.

noyesnoyesyesnoOne prefix or container: volume V, object count N,remaining lifetime L, monthly reads RL longer than the target'sminimum duration?Keep the class: an early movepays the prorated remainderR known and stable?Intelligent-Tiering or smart tier,modelled with N: a monitoringfee per object, no move feesSaving per GB × L above thetransition cost for N objects?A fixed lifecycle ruleNo move; pack small objectsinto archives first
Whether a rule pays off per prefix: the lifetime must outlast the target's minimum duration, the reads must be known, and the per-GB saving over the lifetime must beat the per-object transition cost

Finding candidates

  1. S3 Storage Lens free metrics show NonCurrentVersionStorageBytes and IncompleteMPUStorageBytesOlderThan7Days per bucket. Free metrics stay queryable for 14 days.
  2. S3 Inventory writes daily or weekly CSV, ORC or Parquet with size, last modified date and storage class per object. Query it in Athena for volume, object count and the share under 128 KB per prefix. The first report can take up to 48 hours.
  3. Storage Class Analysis observes access per filter for at least 30 days, but only recommends moving from Standard to Standard-IA, and it is charged separately.
  4. Azure Blob Inventory produces daily or weekly CSV or Parquet with Content-Length, AccessTier, Last-Modified, LastAccessTime, VersionId and IsCurrentVersion, billed per million objects scanned.
  5. Azure access tracking is required for daysAfterLastAccessTimeGreaterThan. Each last-access-time update is billed as an "other transaction", at most once per object every 24 hours.

Example rules for S3 and Azure

The S3 rule tiers logs/ in steps, deletes after the retention period, caps old versions, and cleans up incomplete multipart uploads and orphaned delete markers. The day values are placeholders for your retention rules; each step must satisfy the previous class's minimum duration.

day 30day 120day 400Day 0: written under logs/, Standard,no minimum durationStandard-IA: minimum duration 30 days,so the next step comes after at least 30 days hereGlacier Instant Retrieval: minimum 90 days,so expiration comes after at least 90 days hereExpired
The example rule as a timeline: each step waits out the previous class's minimum duration before the next move or the expiration
{
  "Rules": [
    {
      "ID": "logs-tiering",
      "Status": "Enabled",
      "Filter": { "Prefix": "logs/" },
      "Transitions": [
        { "Days": 30, "StorageClass": "STANDARD_IA" },
        { "Days": 120, "StorageClass": "GLACIER_IR" }
      ],
      "Expiration": { "Days": 400 },
      "NoncurrentVersionExpiration": { "NoncurrentDays": 30, "NewerNoncurrentVersions": 3 }
    },
    {
      "ID": "bucket-hygiene",
      "Status": "Enabled",
      "Filter": { "Prefix": "" },
      "AbortIncompleteMultipartUpload": { "DaysAfterInitiation": 7 },
      "Expiration": { "ExpiredObjectDeleteMarker": true }
    }
  ]
}

Apply it with aws s3api put-bucket-lifecycle-configuration --bucket {bucket} --lifecycle-configuration file://lifecycle.json. The Azure policy does the same for the logs container and deletes previous versions:

{
  "rules": [
    {
      "name": "logs-tiering",
      "enabled": true,
      "type": "Lifecycle",
      "definition": {
        "filters": { "blobTypes": [ "blockBlob" ], "prefixMatch": [ "logs/" ] },
        "actions": {
          "baseBlob": {
            "tierToCool": { "daysAfterModificationGreaterThan": 30 },
            "tierToCold": { "daysAfterModificationGreaterThan": 120 },
            "delete": { "daysAfterModificationGreaterThan": 400 }
          },
          "version": { "delete": { "daysAfterCreationGreaterThan": 30 } }
        }
      }
    }
  ]
}

Azure lifecycle policies themselves are free; you pay for the Set Blob Tier operations they trigger, and deletes cost nothing. New or changed rules can take up to 24 hours to take effect.

Hidden costs: versions and incomplete uploads

In S3, every version is a complete object and is billed in full. Without NoncurrentVersionExpiration, a versioned bucket grows with every overwrite. S3 bills the parts of an unfinished multipart upload until you complete or abort it; removing them with AbortIncompleteMultipartUpload incurs no early deletion charge.

In Azure, deleting a blob with versioning enabled frees no space: the current version becomes a previous version and keeps costing money. Previous versions and soft-deleted data are billed at the same rate as active data. If you set a version's tier explicitly, Azure bills the whole object instead of only its unique blocks.

Common mistakes

  • Archiving incident data. Logs moved to Flexible Retrieval or Archive take hours to read when you need them. Test a restore before widening the rule.
  • Two rules with a short gap. A transition to Glacier Instant Retrieval followed shortly by deletion still costs 90 days.
  • Automatic promotion. enableAutoTierToHotFromCool moves a blob back to Hot on access; if that happens within 30 days, an early deletion fee applies.
  • Re-archiving after rehydration. Without daysAfterLastTierChangeGreaterThan, tierToArchive can archive a freshly rehydrated blob again.
  • Measuring only the storage line. After the change, compare storage, transition, retrieval, request and transfer charges together; reads into another Region add data transfer costs. The individual charge lines come from your cost export, see AWS CUR 2.0 or FOCUS 1.2.

Copies of production data in staging and test belong in the same review; see Non-production cloud costs: control staging, dev and preview. Give every rule an owner, a retention reason and a review date.

Storage costs in Costfluent

Costfluent reads AWS and Azure cost through read-only native connectors. Cost insights highlight significant changes and growing services over the comparison period and lead to the cost rows behind them; in a saved cost report you can filter by resource, such as a bucket or storage account. Cost insights are described in the docs under Insights. That lets you check after a lifecycle change whether storage cost actually falls (for the term, see cost insight in the glossary). You still set lifecycle rules yourself; Costfluent does not change storage configuration.

Sources and method

Checked on 2026-10-10. Amazon S3 storage classes, S3 Lifecycle: transition considerations, S3 Intelligent-Tiering, S3 Lifecycle, S3 Lifecycle examples, Incomplete multipart uploads, Multipart upload pricing, S3 Versioning, Archive retrieval options, Restoring archived objects, Storage Class Analysis, S3 Storage Lens metrics glossary, S3 Inventory, Amazon S3 pricing, AWS Price List API AmazonS3 eu-central-1 (published 28 September 2026) and AmazonS3GlacierDeepArchive eu-central-1 (11 September 2026); Azure access tiers, Smart tier, Access tier best practices, Archive rehydration, Lifecycle management overview, Lifecycle policy structure, Lifecycle policy examples, Blob inventory, Blob versioning, Blob soft delete, Azure Retail Prices API (query for Storage, General Block Blob v2, germanywestcentral, EUR). The tables, formula and review steps are Costfluent recommendations.