Cost-optimized block, file, backup and archive storage
7 min read · about 55 min with practice3 quick checks≈2% of the testCore: Core: tested on most papers
Reading is free. Sign in to tick off lessons, keep your place and track your mastery.
Not every storage bill in Domain 4 comes from S3. The exam also tests waste in block and file storage: volumes sized for IOPS instead of capacity, forgotten volumes and snapshots, cold data kept at hot prices, and backups in the wrong tier. The qualifier is almost always MOST cost-effective, and the answer is usually a managed feature (Elastic Volumes, lifecycle policies, Data Lifecycle Manager, AWS Backup), not a script.
By the end you’ll be able to
Reduce EBS cost by moving gp2 to gp3, using st1/sc1 for throughput-oriented cold data and deleting unattached volumes
Manage snapshot cost with Data Lifecycle Manager, incremental snapshots and the snapshot archive tier
Apply EFS lifecycle management and One Zone storage; choose FSx single-AZ versus multi-AZ and HDD versus SSD
Select the lowest-cost backup/archive solution (AWS Backup tiers, Glacier Deep Archive, Tape Gateway) that meets retrieval needs
Determine the correct storage size and when to enable storage auto scaling
What the exam asks
Choose the cheapest EBS volume type that meets the IOPS and throughput needs, and move gp2 to gp3.
Manage snapshots with Data Lifecycle Manager (DLM) and the EBS Snapshots Archive tier.
Use EFS lifecycle and One Zone, and choose the right FSx deployment and storage type.
Pick the lowest-cost backup and archive design (AWS Backup cold storage, Tape Gateway, Deep Archive) that still meets the restore time, and estimate how long a data transfer will take.
Core ideas
EBS volume types by cost
Type
What you pay for
Use it for
Cannot
gp3 (SSD)
Capacity. 3,000 IOPS and 125 MiB/s are included at any size; more can be added separately
Default for boot volumes, databases, most workloads
—
gp2 (SSD)
Capacity. IOPS = 3 per GiB, and small volumes depend on burst credits
Legacy; migrate to gp3
Get IOPS without buying more GiB
io2 Block Express (SSD)
Capacity + provisioned IOPS
Sub-millisecond, critical databases, Multi-Attach
Be cheap
st1 (HDD)
Capacity (low)
Frequently accessed sequential data: big data, log processing
Boot a system; handle small random I/O
sc1 (HDD)
Capacity (lowest)
Infrequently accessed sequential cold data
Boot a system; handle small random I/O
gp2 → gp3 is the classic quick win: a lower price per GB and IOPS that don’t depend on size. makes the change online, with no downtime.
vii.Check your understanding
3 questions on cost-optimized block, file, backup and archive storage. Every option is explained once you answer.
Sign in to try the quick check
Answers are checked on our side, every option is explained, and your result feeds your mastery for this topic. It’s free.
The first 3 of 10 cards for this topic. Sign in and finish the lesson to review them with spaced repetition.
PromptCard 1 of 3
Why migrate EBS gp2 to gp3?
Elastic Volumes
Elastic Volumes can grow a volume and change its type, IOPS and throughput, but it cannot shrink it. To downsize, create a smaller volume, copy the data and swap. A volume restored from a snapshot must be at least the original volume’s size.
Unattached volumes are billed in full. The root volume is deleted on termination by default. Any volume whose DeleteOnTermination flag is false survives termination, so set the flag in the launch template for scratch volumes.
Snapshots
EBS snapshots are incremental. DLM automates creating, keeping, copying and deleting snapshots on schedules based on tags, with up to four schedules per policy (for example daily kept 14 days, and monthly kept 7 years).
The EBS Snapshots Archive tier costs up to 75% less, but an archived snapshot becomes a full copy, is billed for at least 90 days, and takes up to 72 hours to restore. Archive monthly, yearly or end-of-project snapshots, never short-lived daily ones. DLM can archive snapshots from a monthly or yearly schedule. Fast snapshot restore and the Recycle Bin both cost extra, so enable them only when a requirement asks for it.
Amazon EFS
Lifecycle policies move files to Infrequent Access or Archive based on how long since they were last accessed. An optional policy moves a file back to Standard on first access. The files stay in the same paths, so no application changes are needed.
One Zone file systems cost significantly less than Regional ones. Use them for dev/test data or data that can be re-created. Regional vs One Zone is chosen at creation; to switch, create a new file system and migrate the data with DataSync.
Throughput: Elastic (you pay per use, suits spiky load), Provisioned (steady high throughput), Bursting (scales with the amount stored). Don’t provision throughput you don’t use.
Amazon FSx
Service
Cost levers
FSx for Windows File Server
Single-AZ vs Multi-AZ (only when the share must survive an AZ failure); HDD for general shares and home directories; data deduplication
FSx for Lustre
Scratch (temporary, no replication, cheapest) vs persistent; link to S3, export the results, delete the file system after the job
FSx for NetApp ONTAP
SSD tier sized for active data plus automatic tiering to the low-cost capacity pool; NFS + SMB + iSCSI; deduplication and compression
For most FSx types you can increase SSD capacity but not decrease it. Where a service offers storage auto scaling, as Amazon RDS does, provision for today’s need and set a maximum instead of guessing three years ahead.
Backup, archive and transfer
AWS Backup can move backups of supported resources (including EBS, EFS and DynamoDB with advanced features) to cold storage. There they must stay at least 90 days. Add cross-Region or cross-account copies only when a requirement asks for them.
Tape Gateway presents a virtual tape library over iSCSI, so the backup software doesn’t change. Ejected tapes are archived to Glacier Flexible Retrieval or Deep Archive.
Data transfer into AWS is free. Transfer time is data ÷ bandwidth: 60 TB = 480,000 Gb, which takes about 5.6 days at 1 Gbps. If the link meets the deadline, use AWS DataSync and write straight into the final S3 storage class. Don’t buy Direct Connect for a one-time copy. (Snowball still appears in the SAA-C03 service list, but AWS ends Snowball device support on 31 December 2026.)
Worked examples
Exam technique
“Burst credits”, “gp2”, “IOPS independent of size” → gp3.
“Sequential”, “infrequently accessed”, “lowest cost”, “not a boot volume” → sc1; if frequently accessed → st1.
“Not accessed after N days but must stay in the file system” → EFS lifecycle.
“Can be re-created”, “dev/test” → One Zone.
“Temporary high-performance processing of S3 data” → FSx for Lustre scratch.
“Existing backup software writes to tape” → Tape Gateway; “48 hours to restore” → Deep Archive.
Snapshots kept less than 90 days → standard tier. Kept for years and rarely restored → archive tier.
Common mistakes
Quick recap
gp3 gives 3,000 IOPS and 125 MiB/s at any size for less than gp2. Change type with Elastic Volumes, online.
sc1 is for cold sequential data and st1 for hot sequential data. Neither can boot a system.
Delete unattached volumes and set DeleteOnTermination on scratch volumes. Volumes never shrink in place.
DLM for snapshot schedules. The archive tier (full copy, 90-day minimum, restore in up to 72 hours) is for long-kept snapshots only.
EFS: lifecycle to IA or Archive, One Zone for re-creatable data.
FSx: Multi-AZ only when required, HDD for general shares, Lustre scratch for temporary jobs, ONTAP capacity pool tiering.
AWS Backup cold storage needs at least 90 days there. Tape Gateway plus Deep Archive replaces physical tape.