Storage
Amazon EBS
Block storage for Amazon EC2 instances: volumes that you attach to an instance and use like a local hard drive, plus snapshots that store point-in-time copies. It provides an instance's disk, for example to store files or install applications.
Exam phrases that point here
«persistent block storage for an EC2 instance»
An Amazon EBS volume behaves like a disk attached to the instance, and its data survives even if the instance is stopped or terminated (if so configured).
When not: If several instances, even in different Availability Zones, must share the same files, Amazon EFS fits. If the data is temporary and can be lost along with the instance, instance store may be enough.
«mission-critical database that needs high, consistent IOPS with very low latency»
io2 Block Express volumes are designed for demanding databases: sub-millisecond latency, provisioned IOPS, and higher durability than the other EBS volumes.
When not: If a balance between price and performance is enough, gp3 is usually the choice. And if the scenario prioritizes not managing the database, the answer may be a managed service such as Amazon RDS instead of choosing the volume.
«high-throughput sequential workloads, such as big data or log processing»
HDD volumes (st1 and sc1) are optimized for large, sequential reads and writes, measured in MB/s, at a lower cost than SSDs.
When not: With small, random I/O, such as that of a transactional database, HDDs perform poorly and the answer becomes an SSD volume (gp3 or io2). Of the two HDDs, sc1 is for cold data that is accessed infrequently.
«automate the creation, retention, and deletion of EBS snapshots»
Amazon Data Lifecycle Manager creates snapshots of volumes or instances with certain tags on a schedule, deletes old ones according to retention, and can copy them to other Regions, at no additional cost.
When not: If you need to manage backups of several services (EBS, RDS, DynamoDB, EFS...) from a single place, or apply policies across the whole organization, the tool is AWS Backup.
«attach the same volume to multiple instances in a cluster»
EBS Multi-Attach lets you attach a Provisioned IOPS volume (io1 or io2) to multiple instances in the same Availability Zone at once, for cluster applications that coordinate access to the disk.
When not: The application must coordinate writes (for example, Windows Server Failover Cluster or NVMe reservations); a regular file system would be corrupted. If all that is needed is to share files between instances, the answer is Amazon EFS. A volume with Multi-Attach cannot be used as a boot volume.
«ensure that all new volumes are encrypted»
EBS encryption by default is turned on with a single setting per Region and requires every new volume to be created encrypted with AWS KMS.
When not: The setting is per Region: it must be turned on in each Region you use. If you must also share encrypted snapshots with other accounts, they must be encrypted with a customer managed key, not the default key.