Storage
Storage is the part of a computing system that retains data so applications can read, modify and reuse it.
Storage design decides where data lives, how it is accessed, how long it must surviveand how quickly it can be recovered.
RAM and GPU memory hold working data for active computation. Persistent storage retains saved data beyond a process or machine restart, subject to the service and configuration. SSDs use flash; HDDs use magnetic disks. These are storage media, while block, file and object
describe access models. A local SSD can still be temporary if its cloud resource lifecycle removes the data.
Block, file and object
These three models describe how applications address data. Block storage exposes numbered blocks; file storage exposes files and directories; object storage exposes named objects through an API. They are not competing levels of quality. A single application can use all three because its database, shared workspace and document archive have different access needs.
Capacity and performance
Capacity is how much data fits. Latency is the delay for an operation. IOPS counts operations per second, while throughput measures bytes moved per second. Random access jumps between locations; sequential access reads or writes contiguous data. The right metric depends on whether the workload handles small database updates, large video files or batches of training examples.
Durability vs Availability
Durability describes resistance to data loss. Availability describes whether data can be accessed when needed. Consistency describes which version a reader sees after a write. A storage service can retain every byte during an outage and still be temporarily unavailable. Replication, backups and access semantics solve different problems and should be evaluated separately.
Storage on AWS
Object Storage
Amazon Simple Storage Service (S3) stores data as objects addressed through APIs. In a general purpose bucket, an object has a key, data and metadata. A key such as lessons/storage.pdf is a name; the slash-delimited prefix can look like folders without being a traditional file-system directory. S3 is widely used for datasets, media, logs, backups and model artifact
Applications use operations such as PUT, GET and LIST through an SDK, CLI or other client. For
ordinary general purpose objects, replacing content writes a new object rather than editing disk blocks in place. Parallel requests and suitable upload patterns improve large-data workflows. S3 is a strong fit for a durable data repository, but software expecting complete POSIX file behavior needs a compatible access layer.
Block Storage
Amazon Elastic Block Store (EBS) provides persistent block volumes for EC2. A volume is network-attached storage in an Availability Zone and normally attaches to an instance in that same AZ. You can create a file system on it or let a database use its blocks. The volume can outlive an instance, depending on its deletion settings.
File Storage
Amazon Elastic File System (EFS) is managed, elastic file storage accessed through NFS by compatible clients, commonly Linux workloads. Multiple clients can use the same file system. This suits shared application files, home directories and tools that expect file paths. EFS handles storage infrastructure while the application still needs appropriate file permissions and concurrent-access behavior.
