High-Performing and Scalable Storage

Bundled into the existing AWS Certified Solutions Architect – Associate premium course — no separate purchase.

14-day money-back guarantee — no questions asked.

Included in this chapter:

Type	Use case	Max IOPS / volume	Max throughput / volume	Durability	Cost
gp3 (default)	Most general-purpose workloads	80 000 (provisioned)	2 000 MB/s (provisioned)	99.8–99.9%	$ (cheapest SSD; baseline + provisioned)
gp2 (legacy)	Migrate to gp3	16 000 (size-tied)	250 MB/s	99.8–99.9%	$ (slightly more than gp3)
io2 Block Express (every io2 volume)	Mission-critical DB > 80,000 IOPS, consistent sub-ms latency	256 000 (Nitro-based instances; 32 000 on others)	4 000 MB/s	99.999%	$$$$
io1 (previous generation)	High IOPS DB	64 000	1 000 MB/s	99.8–99.9%	$$$
st1	Big-data sequential, log processing	500	500 MB/s	99.8%	$ (HDD)
sc1	Rarely-accessed cold block storage	250	250 MB/s	99.8%	$ (HDD, lowest)

Cheat sheet

Pick storage by how it's accessed: object, file, or block
Match the EBS volume type to whatever the workload is bound on
EFS performance mode and throughput mode are tuned separately
gp3 is the default, and you can migrate gp2 in place to save ~20%
For >80,000 IOPS or 99.999% durability, reach for io2 Block Express
Instance store is free but vanishes on stop, terminate, or hibernate
FSx for Lustre gives you HPC scratch at TB/s-class throughput, linked to S3
EFS Bursting banks throughput credits while idle and spends them under load
EBS Multi-Attach works only on io1/io2, and only with a cluster-aware filesystem
For AD-integrated SMB shares, use FSx for Windows or NetApp ONTAP
When the access pattern is unknown or shifting, use S3 Intelligent-Tiering
For cold archive that still needs millisecond access, use Glacier Instant Retrieval
S3 Transfer Acceleration needs a dot-free, DNS-compliant bucket name

Unlock with Premium — includes all practice exams and the complete study guide.

Cheat sheet