Skip to content
Certification Guide

Task 4.1 · 24 min read

Cost-Optimized Storage

Services in this lesson

  • Amazon S3in scope
  • Amazon S3 Glacierin scope
  • Amazon EBSin scope
  • Amazon EFSin scope
  • Amazon FSx (for all types)in scope
  • AWS Backupin scope
  • AWS Storage Gatewayin scope
  • AWS DataSyncin scope
  • AWS Transfer Familyin scope
  • AWS Cost Explorerin scope
  • AWS Budgetsin scope
  • AWS Cost and Usage Reportin scope

What the exam tests

  • Picking the cheapest S3 (Amazon Simple Storage Service) storage class that still meets the stated read frequency, retrieval time and resilience. Spotting when a minimum storage duration, minimum billable size or retrieval fee erases the savings.
  • Writing S3 Lifecycle rules that transition, expire and clean up data (old versions, incomplete multipart uploads) instead of relying on people or scripts.
  • Choosing the lower-cost option inside EBS, EFS and FSx. That means gp3 over gp2, HDD volumes for sequential or cold data, and EFS lifecycle and One Zone. On FSx, it means HDD storage, deduplication (storing repeated data only once) and scratch file systems (temporary and not replicated). It also means snapshot and backup cold tiers.
  • Choosing the lowest-cost way to move data into AWS storage, and the right hybrid service (AWS DataSync, AWS Transfer Family, AWS Storage Gateway), when cost decides.
  • Using Requester Pays, cost allocation tags, AWS Cost Explorer, AWS Budgets and the AWS Cost and Usage Report to see and control storage spend. With Requester Pays, the requester, not the bucket owner, pays request and download charges.

The core concept

Storage cost questions have one shape: the same data in a cheaper place. Every cheaper place charges you elsewhere, through a retrieval fee, restore delay, minimum storage duration, minimum object size or less resilience.

An analogy: a self-storage company

A self-storage company rents space at several prices. The locker by the front desk is the priciest, but you can open it any time. The units at the back are cheaper, but staff charge a small fee each time they fetch something.

The off-site depot is the cheapest of all. Items take hours to come back, and you must rent the space for at least six months. Moving a box out early still costs you the full term.

Where it breaks: a real depot can bring your box back to the front-desk locker on request. In S3, lifecycle rules only move data down the waterfall toward colder classes, never back up. You move an object up yourself by restoring and copying it.

Walk one archive through it

The diagram is the S3 lifecycle waterfall. The walkthrough follows one company's data down it.

Enlarge
The S3 lifecycle waterfall: each step down is cheaper to store and costlier or slower to read.

Juniper Archive scans paper records for law firms and keeps the scans in S3. Noor is its cloud architect, and Henrik runs finance. Noor asks three questions of every dataset: how often is it read, how fast must it come back, and can it be re-created?

Data enters the diagram at one of two points. Steps 1 and 2 are two separate datasets. Steps 3 to 6 follow the second one, the scans, down the waterfall.

  1. Dataset one: customer uploads arrive with no known pattern. Some files are read daily, others never. Noor sends them to S3 Intelligent-Tiering, the class that moves each object between access tiers by itself.
    • After 30 days without access, an object moves to the Infrequent Access tier. After 90 days, it moves to Archive Instant Access.
    • Intelligent-Tiering charges a small monitoring fee per object and no retrieval fee.
  2. Dataset two: fresh scans have a known pattern. Lawyers read them heavily for a month, so they start in S3 Standard, the default class with no minimum duration.
  3. After 30 days, reads drop to about once a month. A lifecycle rule (a bucket setting that moves or deletes objects on a schedule) moves them to S3 Standard-IA (Infrequent Access).
    • Standard-IA is cheaper to store but charges a per-GB retrieval fee.
    • It bills each object for at least 30 days and at least 128 KB.
  4. After 90 days, reads drop to about once a quarter. The rule moves scans to S3 Glacier Instant Retrieval, which still returns data in milliseconds. Its minimum storage duration is 90 days.
  5. After six months, scans are kept only for audit. The rule moves them to S3 Glacier Deep Archive, the cheapest class. Its minimum is 180 days. S3 Glacier Flexible Retrieval, the class just before it, has a 90-day minimum.
    • Objects in these two classes must be restored (copied back temporarily) before anyone can read them.
    • To move one back to a warmer class, Noor must restore it and then copy it.
  6. After seven years, the scans expire. The lifecycle rule deletes them, so nobody has to remember.
  7. Henrik asks which team drives the bill. Noor activates cost allocation tags (labels on resources that split spend by team or project). Henrik then views spend by tag in AWS Cost Explorer.

Each step down saves storage money only because Noor matched it to how the data is really read.

The rules the exam tests

  • Minimum storage durations: 30 days for S3 Standard-IA and S3 One Zone-IA. 90 days for S3 Glacier Instant Retrieval and S3 Glacier Flexible Retrieval. 180 days for S3 Glacier Deep Archive.
  • Early exit still pays. Delete, overwrite or transition an object before its minimum, and you pay for the rest of that minimum.
  • Minimum billable size: S3 Standard-IA, S3 One Zone-IA and S3 Glacier Instant Retrieval bill at least 128 KB per object.
  • Unknown or changing pattern: S3 Intelligent-Tiering.
  • Known pattern, unknown transition day: turn on S3 Storage Class Analysis. It watches access for a group of objects and says when they are cold enough to move from S3 Standard to S3 Standard-IA. That is the number your lifecycle rule needs.
  • Lifecycle only moves data down the waterfall. Leaving S3 Glacier Flexible Retrieval or S3 Glacier Deep Archive for a warmer class takes a restore and then a copy.
Deep diveSmall objects: 128 KB and 40 KB

Three small-object rules use similar numbers, so they are easy to mix up.

First, S3 Standard-IA, S3 One Zone-IA and S3 Glacier Instant Retrieval bill every object as at least 128 KB. A 16 KB object there costs as much to store as a 128 KB one.

Second, S3 Lifecycle does not transition objects under 128 KB by default. The per-object transition request could cost more than the storage saved.

Third, S3 Glacier Flexible Retrieval and S3 Glacier Deep Archive have no minimum size. Instead, each archived object carries about 40 KB of extra billed metadata. Bundle tiny objects into larger files before archiving them.

Deep diveIntelligent-Tiering limits

Objects under 128 KB are not monitored. They always stay in the Frequent Access tier, so Intelligent-Tiering saves nothing on tiny objects.

The optional Archive Access and Deep Archive Access tiers need a restore before a read. Turn them on only when the application can wait.

This lesson only asks which option costs least while meeting every requirement. Speed and throughput belong to task 3.1, storage performance. Encryption, Object Lock and replication for protection belong to task 1.3, data security.

Noor's walkthrough stayed in S3. Juniper Archive's servers and shared drives follow the same idea with different levers.

Block and file storage

Amazon EBS (Elastic Block Store) provides network disks for one Amazon EC2 (Elastic Compute Cloud) instance, a virtual server. The classic EBS saving is moving from gp2 to gp3. Both are SSD (solid-state drive) volume types.

  • gp3 includes 3,000 IOPS (input/output operations per second) and 125 MiB/s (mebibytes per second) in the storage price.
  • gp3 costs up to 20 percent less per GiB (gibibyte) than gp2.
  • Elastic Volumes, the feature that changes a live volume, switches gp2 to gp3, usually with no downtime.
  • HDD (hard disk drive) types, st1 and sc1, are cheaper per GiB for large sequential or cold data. Neither can be a boot volume.

Amazon EFS (Elastic File System) is a shared Linux file system. Amazon FSx runs managed file systems: FSx for Windows File Server, FSx for Lustre (high-performance computing) and FSx for NetApp ONTAP (multiprotocol NetApp storage). On both, you save three ways:

  • Tier cold files to cheaper storage, such as EFS Infrequent Access or the ONTAP capacity pool (an elastic, lower-cost tier).
  • Give up Multi-AZ (copies in several Availability Zones) where data can be re-created.
  • Choose HDD storage or a scratch file system (temporary, not replicated) when performance allows.

Sizing is a cost decision too.

S3 and EFS grow and shrink with your data, so you pay for what you store. EBS volumes and FSx file systems bill for the capacity you provision. Size them for real use plus headroom, and grow them when monitoring says so.

An EBS volume can grow with Elastic Volumes but can never shrink. Over-provisioning is a cost you keep. Database storage autoscaling belongs to task 4.3.

Seeing where the money goes

Henrik's question needs the billing tools, each for its own job. First, activate cost allocation tags so spend can be split by team or project.

  • AWS Cost Explorer explores spend in the console.
  • AWS Budgets alerts on actual or forecasted spend.
  • AWS Cost and Usage Report gives line-item billing data. AWS delivers it to an S3 bucket you own, and you typically query it with Amazon Athena (SQL over files in S3).
Deep diveConsolidated billing in AWS Organizations

AWS Organizations manages many AWS accounts centrally. Its consolidated billing combines usage across member accounts. Volume pricing discounts then apply to the organization's total.

Decision guide

Decision guide
If the question says…Reach for…Because…
"unknown", "unpredictable" or "changing" access pattern, no operational overheadS3 Intelligent-TieringMoves each object between tiers by itself. No retrieval fees, only a small per-object monitoring fee
read about once a month, must be available in milliseconds, only copy of the dataS3 Standard-IASame multi-AZ resilience as S3 Standard, cheaper to store. Adds a per-GB retrieval fee and 30-day minimum
infrequent reads, data can be re-created (thumbnails, replicas, derived files)S3 One Zone-IACheaper than S3 Standard-IA because it uses one Availability Zone. Lose that zone and you lose the data
read about once a quarter, still needs millisecond accessS3 Glacier Instant RetrievalArchive-level storage price with immediate reads. 90-day minimum and higher retrieval fees
archive, retrieval in minutes to hours is acceptableS3 Glacier Flexible RetrievalRestore first: Expedited 1–5 minutes (under 250 MB), Standard 3–5 hours, Bulk 5–12 hours. 90-day minimum
compliance archive kept for years, retrieval within 12–48 hours is fineS3 Glacier Deep ArchiveLowest S3 storage price. Standard restores finish within 12 hours, Bulk within 48. 180-day minimum
known pattern, but which day should data move from S3 Standard to S3 Standard-IA?S3 Storage Class AnalysisObserves access for a prefix or tag group, then recommends when to transition to S3 Standard-IA
bucket keeps growing with old versions or half-finished uploadsLifecycle NoncurrentVersionExpiration and AbortIncompleteMultipartUploadOld versions and incomplete upload parts are billed as storage until a rule removes them
share a large dataset with other AWS accounts without paying for their downloadsS3 Requester PaysThe requester pays request and data transfer charges. The bucket owner still pays for storage
gp2 volumes, cut cost without downtimeChange to gp3 with Elastic VolumesUp to 20 percent cheaper per GiB than gp2. Includes 3,000 IOPS and 125 MiB/s at any size
large sequential reads (logs, big data) on block storage, lowest costst1 (or sc1 if the data is rarely read)HDD types cost less per GiB and are built for throughput. Neither can be a boot volume
monthly or yearly EBS snapshots kept for years, rarely restoredEBS Snapshots ArchiveLower price for full snapshots. 90-day minimum, and restores take up to 72 hours
shared Linux files, most untouched after a few weeksEFS lifecycle management to Infrequent Access and ArchiveFiles move by last access time. IA and Archive cost less to store but charge per read
Windows home directories or departmental shares, cost firstFSx for Windows File Server with HDD storage and data deduplicationHDD suits general shares. AWS cites average deduplication savings of 50–60 percent on general file shares
FSx for NetApp ONTAP volumes where only a small part of the data is activeAuto tiering policy to the capacity poolCold data moves to the elastic, lower-cost capacity pool after a cooling period. Active data stays on SSD
short-lived HPC job reading from S3FSx for Lustre scratch file system linked to the bucketScratch is the cost-optimized choice for temporary processing. Data isn't replicated, so S3 stays durable
keep backups of several services for years at low costAWS Backup lifecycle to cold storage (EBS Snapshots Archive for one volume's monthly snapshots)One plan moves supported backups to cold storage after a warm period. Cold storage has a 90-day minimum
replace a physical tape library, backup software stays the sameStorage Gateway Tape Gateway archiving to S3 Glacier Deep ArchiveVirtual tapes look like tape to the backup software and archive into the cheapest S3 class
one-time or scheduled copy from on-premises NFS or SMB over the existing linkAWS DataSyncManaged transfer with validation and metadata preservation. You pay only for usage and maintain no scripts
partners must upload over SFTP or FTPS, no servers to runAWS Transfer FamilyA fully managed SFTP, FTPS, FTP or AS2 endpoint that writes straight into S3 or EFS
the network would take weeks to move the dataAn offline transfer device (AWS Snow Family, existing customers)Shipping disks beats a slow link. AWS no longer offers Snow Family devices to new customers
see which team's buckets drive the bill; alert before spend overshootsCost allocation tags + Cost Explorer; AWS Budgets forecast alertsUser-defined tags show in Cost Explorer only once activated. Budgets alerts on actual or forecasted spend
analyze line-item storage charges with SQL across monthsAWS Cost and Usage Report delivered to S3, queried with Amazon AthenaCost Explorer explores, Budgets alerts, and the Cost and Usage Report analyzes

A few terms from the table: HPC is high-performance computing. NFS (Network File System) and SMB (Server Message Block) are file-sharing protocols. SFTP, FTPS, FTP and AS2 are file-transfer protocols that partners commonly use.

Tape Gateway is the AWS Storage Gateway type that presents virtual tapes to existing backup software.

Correct pattern

REFERENCE PATTERN: one S3 Lifecycle configuration for an application log bucket. Logs are read heavily for a month, occasionally for a quarter, and kept for about seven years for audit. The bucket is versioned, so overwrites keep the old copy as a noncurrent version.

{
  "Rules": [
    {
      "ID": "logs-cool-down-and-expire",
      "Status": "Enabled",
      "Filter": { "Prefix": "logs/" },
      "Transitions": [
        { "Days": 30, "StorageClass": "STANDARD_IA" },
        { "Days": 90, "StorageClass": "GLACIER_IR" },
        { "Days": 180, "StorageClass": "DEEP_ARCHIVE" }
      ],
      "Expiration": { "Days": 2557 },
      "NoncurrentVersionExpiration": { "NoncurrentDays": 30 }
    },
    {
      "ID": "clean-up-incomplete-uploads",
      "Status": "Enabled",
      "Filter": {},
      "AbortIncompleteMultipartUpload": { "DaysAfterInitiation": 7 }
    }
  ]
}

Why it holds up:

  • Each transition leaves the object in the previous class for at least that class's minimum. Objects spend 60 days in S3 Standard-IA (minimum 30) and 90 days in S3 Glacier Instant Retrieval (minimum 90). A rule that skips this is rejected or pays early-deletion charges.
  • Expiration at 2,557 days covers seven years including leap days, so nobody has to remember to delete.
  • In a versioned bucket, Expiration only adds a delete marker (a placeholder that hides the object). The log becomes a noncurrent version. NoncurrentVersionExpiration is what permanently removes the data 30 days later.
  • A multipart upload sends a large file in parts. Abandoned parts and old versions of overwritten logs would otherwise be billed forever.
  • By default, lifecycle does not transition objects under 128 KB. The per-object transition request can cost more than the storage saved. If your logs are tiny, combine them into larger files before upload.

The block storage equivalent is one call. On current-generation instances it usually runs without detaching the volume or stopping the instance:

aws ec2 modify-volume --volume-id <vol-id> --volume-type gp3

Traps

TRAPExam traps
The trapWhy it failsCorrect pattern
Lowest storage price wins: put the archive in S3 Glacier Deep ArchiveCheapest class tempts. Reads needed in minutes, or every week, hit restore delays or retrieval feesRead the retrieval time first. Milliseconds: S3 Glacier Instant Retrieval. Minutes: S3 Glacier Flexible with Expedited
S3 One Zone-IA for the only copy of business recordsTempting: cheaper than S3 Standard-IA with the same millisecond access. One zone lost can destroy the dataS3 One Zone-IA only for re-creatable data or replicas. S3 Standard-IA for the only copy
Lifecycle millions of small objects straight to an archive classLooks like a free saving. Transitions are billed, objects under 128 KB don't move by default, archives add 40 KBBundle small objects into larger archives first, or keep them in a class without per-object minimums
Move short-lived data (deleted after two weeks) to S3 Standard-IA to save moneyLower per-GB price tempts. The 30-day minimum bills 30 days anyway, plus retrieval feesKeep short-lived data in S3 Standard and expire it with a lifecycle rule
Turn on S3 Intelligent-Tiering Archive Access for data the application reads directlySounds like more savings. Archive tier objects need a restore before a GET, so application reads failLeave the optional archive tiers off unless the application can wait for a restore
Shrink an over-provisioned EBS volume with Elastic VolumesFeels like the natural undo for over-sizing. Elastic Volumes can grow a volume or change its type, never shrink itCreate a smaller volume, copy the data, swap it in. Or change gp2 to gp3 to cut the per-GiB price
Use sc1 for the boot volume because it is the cheapest EBS typeCheapest per GiB, so it looks right for a boot disk. HDD volume types (st1, sc1) can't be boot volumesgp3 for boot volumes. st1 or sc1 only for large sequential data volumes
AWS Backup plan: move to cold storage after 300 days, delete after 365Looks tidy, but cold storage needs at least 90 days. Retention must be at least the cold-storage day plus 90Move to cold storage earlier (for example after 30 days), or keep short-retention backups in warm storage

A GET is the S3 request that reads an object.

Compare

Compare
AspectS3 Standard-IAS3 One Zone-IAS3 Glacier Instant RetrievalS3 Glacier Flexible RetrievalS3 Glacier Deep ArchiveS3 Intelligent-Tiering
Designed for reads aboutOnce a monthOnce a monthOnce a quarterOnce a yearLess than once a yearUnknown or changing
First byteMillisecondsMillisecondsMillisecondsAfter restore: 1 minute to 12 hoursAfter restore: up to 48 hoursMilliseconds (optional archive tiers need a restore)
Availability Zones3 or more13 or more3 or more3 or more3 or more
Minimum storage duration30 days30 days90 days90 days180 daysNone
Minimum billable object size128 KB128 KB128 KBNone (about 40 KB metadata added)None (about 40 KB metadata added)None (under 128 KB stays in Frequent Access)
Retrieval feePer GBPer GBPer GBPer GB (Bulk is free)Per GBNone; small monitoring fee per object

For block and file storage, the cost levers sit inside each service:

Compare
ServiceCheaper optionWhat you give up
Amazon EBSgp3 instead of gp2; st1 or sc1 for sequential or cold dataNothing for gp3. HDD types can't boot and are slow for random I/O
EBS snapshotsSnapshots Archive for long-term snapshots90-day minimum, up to 72 hours to restore before use
Amazon EFSLifecycle to IA and Archive; One Zone file systemPer-read charge on cold files. One Zone is lost with its zone. Archive needs Regional and Elastic throughput
FSx for Windows File ServerHDD storage; data deduplicationHigher latency than SSD. Dedup jobs use file server resources
FSx for NetApp ONTAPTiering policy (Auto, Snapshot Only or All) to the capacity poolMetadata always stays on SSD. Under Auto, randomly read cold blocks are made hot again
FSx for LustreScratch deploymentNo replication. A failed server means I/O errors on its files
AWS BackupCold storage tier (for supported resources such as Amazon EFS)90-day minimum in cold storage. Slower, costlier restores

On EFS, the Archive storage class needs a Regional file system (data stored across several Availability Zones) and Elastic throughput (a mode that scales throughput with traffic). One Zone file systems can't use it.

AWS Backup first appears in task 1.3, data security, where it protects data. Here it is a cost tool.

Deep diveBackup cold storage and offline transfers

One AWS Backup plan can move supported backups, such as Amazon EFS backups, to cold storage after a warm period. AWS Backup recommends a warm storage duration of at least one week. Cold storage has a 90-day minimum.

For one EBS volume's monthly snapshots, EBS Snapshots Archive is the direct equivalent.

If the network would take weeks, the classic exam answer is an offline AWS Snow Family device. AWS no longer offers these devices to new customers, so treat that answer as legacy and use the transfer table below.

Moving data in has its own cost answers, and each tool fits a different job.

Compare
NeedReach forNote
Copy file or object data over a working network linkAWS DataSyncValidates data and preserves metadata; you pay only for usage
Partners upload by SFTP, FTPS, FTP or AS2 into S3 or EFSAWS Transfer FamilyFully managed endpoint, so you run no servers for it
Ongoing hybrid access to files from on premisesAWS Storage GatewayHybrid access, not a one-time move; mechanics are in task 3.1
A large offline transfer, where the network would take weeksAWS Partner solutions, or DataSync over the network if it can finish in timeAWS no longer offers Snow Family devices to new customers

Data transfer pricing between Regions, Availability Zones and the internet belongs to task 4.4.

Worked exam scenario

Exam scenario

A research institute stores instrument readings in Amazon S3. Each object is about 500 MB. Scientists read a new object many times in its first 30 days, about once a quarter until it is one year old, and almost never after that. Quarterly reads must return data immediately. When a scientist does request an object that is more than one year old, a result within 12 hours is acceptable. Regulations require keeping every object for 10 years. Which solution meets these requirements MOST cost-effectively?

  1. A.Store objects in S3 Intelligent-Tiering with the Archive Access and Deep Archive Access tiers activated. Add a lifecycle rule to expire objects after 10 years.
  2. B.Store objects in S3 Standard. Add a lifecycle rule that transitions objects to S3 Glacier Instant Retrieval after 30 days and to S3 Glacier Deep Archive after 365 days, and expires them after 10 years.
  3. C.Store objects in S3 Standard. Add a lifecycle rule that transitions objects to S3 One Zone-IA after 30 days and to S3 Glacier Flexible Retrieval after 365 days, and expires them after 10 years.
  4. D.Store objects in S3 Standard-IA. Add a lifecycle rule that transitions objects to S3 Glacier Deep Archive after 30 days and expires them after 10 years.
Show the walkthrough
  1. Underline the requirements. Heavy reads for 30 days. Quarterly reads that must return immediately until one year. Twelve hours is fine after that. Keep 10 years, MOST cost-effectively. The access pattern is known, so you can name each tier yourself.
  2. Map each phase to a class. First 30 days: S3 Standard, for frequent reads with no minimum. Days 30–365, read quarterly and returned immediately: S3 Glacier Instant Retrieval. After a year: S3 Glacier Deep Archive, whose Standard restores finish within 12 hours.
  3. Eliminate A. S3 Intelligent-Tiering fits unknown patterns. Here the pattern is known, so the monitoring fee buys nothing. Its Deep Archive Access tier also starts only after 180 days without access, and quarterly reads keep resetting that clock.
  4. Eliminate C. S3 One Zone-IA puts the only copy of regulated data in one Availability Zone. S3 Glacier Flexible Retrieval also costs more to store than Deep Archive when 12 hours is already acceptable.
  5. Eliminate D. Writing new objects to S3 Standard-IA charges a retrieval fee on every read in the heavy first month. Moving to Deep Archive after 30 days breaks the quarterly reads with restores.
  6. Land on B. Every phase sits in the cheapest class that still meets its retrieval need. Each transition respects the earlier class's minimum: 335 days in S3 Glacier Instant Retrieval is well past its 90-day minimum.

Answer: B

Check your understanding

Question 1: A team uploads 2 million objects of about 16 KB each and wants a lifecycle rule to move them to S3 Glacier Flexible Retrieval after 60 days to save money. What happens by default?
Question 2: Which TWO options lower EBS cost without changing the application or stopping the instance? (Choose TWO.) (Choose TWO.)
Question 3: A reporting job writes export files to S3 that are downloaded once and deleted after 10 days. To save money, a developer proposes writing them to S3 Standard-IA instead of S3 Standard. What is the effect?
Question 4: A team runs an EFS One Zone file system in Bursting throughput mode. It wants a lifecycle policy that moves files not read for 180 days into the EFS Archive storage class. What must change before that policy can work?

Build it

Build a lifecycle-managed bucket, try to read an archived object and watch the request fail, then restore it in the cost-optimized storage lab.

Key takeaways

  • Read the scenario for read frequency, retrieval time and whether data can be re-created. Those three facts pick the storage class.
  • Minimum storage durations (30, 90 or 180 days), the 128 KB minimum and per-object charges can make a cheaper class cost more for short-lived or tiny objects.
  • Unknown access pattern means S3 Intelligent-Tiering. A known pattern means a lifecycle rule that transitions, expires and cleans up old versions and incomplete uploads.
  • Outside S3, save with gp3 over gp2, HDD volumes for sequential or cold data, and snapshot and backup cold tiers. On file systems, use EFS lifecycle and One Zone, FSx HDD, deduplication and scratch.
  • Activate cost allocation tags, then use Cost Explorer to explore, AWS Budgets to alert and the Cost and Usage Report to analyze. Storage growth then shows before the bill arrives.
  • Pay for the access pattern you have: match each dataset's read frequency and tolerable retrieval time to the cheapest tier meeting both, then let lifecycle rules move and expire data.

Hands-on lab: Lifecycle a Bucket and Hit the Archive Wall

Sources

Last reviewed