S3 (object storage)
Also known as: Simple Storage Service, S3 bucket, AWS storage
S3 (Simple Storage Service) is AWS's storage bucket service.
You upload any file -- images, videos, backups, logs, datasets -- and S3 stores it durably and makes it accessible via a URL. There's no limit on how much you can store, and you pay per gigabyte.
S3 (object storage), explained at three levels
At home
Think of S3 as a hard drive in the cloud with no size limit. You create a 'bucket' (a named container), drop files in, and they're stored safely across multiple Amazon data centers. Files can be public (like images on a website) or private (like backups).
At work
Businesses use S3 to store everything they can't lose -- database backups, customer uploads, application logs, and media files. S3 automatically replicates your data across multiple facilities so a single data center failure can't destroy it.
At home
S3 organizes files into buckets with key-based addressing (the 'path' is part of the key, not a real folder structure). Storage classes let you trade cost for access speed: S3 Standard for frequent access, S3 Infrequent Access for occasional, and S3 Glacier for long-term archival at a fraction of the cost.
At work
S3 integrates with virtually every AWS service -- Lambda triggers on new uploads, Athena queries CSV/Parquet files directly, CloudFront serves S3 objects as a CDN, and Lifecycle Policies automatically move files to cheaper storage tiers after a set number of days.
At home
S3's consistency model changed in 2020: all operations (including overwrite PUTs and DELETEs) are now strongly consistent, meaning a read immediately after a write sees the latest version. This simplified many distributed application designs.
At work
At scale, S3 supports thousands of requests per second per prefix. For very high-throughput workloads, using multiple key prefixes distributes load. S3 Object Lock provides WORM (write-once-read-many) compliance storage. S3 Replication can replicate objects cross-region or cross-account for disaster recovery or data sovereignty requirements.
Real-world examples
At home
Sophie uses an S3 bucket to host her static portfolio website. She uploads HTML, CSS, and image files, enables static website hosting in the bucket settings, and her site is live -- no web server needed.
At work
Jerome's company stores all database backups in S3 with a lifecycle policy: backups move to S3-IA after 30 days and to Glacier after 90 days. This cut their storage bill by 70% compared to keeping everything in S3 Standard.
Common confusions
- S3 is like a file system with folders — S3 is a flat key-value store. The slashes in object keys look like folder paths, but there are no real directories -- it's a naming convention. This affects how you list and prefix-filter objects.
- S3 is too expensive for large datasets — S3 Standard costs around $0.023/GB/month. For 10 TB that's $23/month. With Intelligent Tiering or Glacier for cold data, costs drop dramatically. It's usually far cheaper than on-premises storage.
For analysts
Data analysts frequently land raw data in S3 then query it in place using Athena (SQL) or load it into Redshift. S3 functions as a data lake foundation -- store everything, query what you need.
History & origin
S3 launched alongside EC2 in March 2006 as one of AWS's founding services. It was designed around a simple idea: a key-value store for objects of any size, infinitely scalable, with 99.999999999% (11 nines) durability. S3 popularized the 'bucket' abstraction and became so dominant that its API is now the industry standard -- dozens of other object storage services (MinIO, Wasabi, Backblaze B2) implement the S3 API for compatibility.