TOPIC CLOUD
Object Storage unlimited, highly durable storage for any blob of data
IN 10 SECONDS
Buckets hold objects (files) at URLs like `s3://my-bucket/path/to/file.jpg`. Each object has data, metadata, and a unique version ID. Storage classes trade cost for access frequency — Standard, Infrequent Access, Glacier (archive). 11 nines of durability means you can lose two datacenters and still read every object.
GOTCHA S3 is eventually consistent for overwrite PUTS and DELETES (unless you use read-after-write consistency for new objects). A `ListObjects` right after a delete might still show the deleted object for a few seconds.
HOW AN S3 GET REQUEST WORKS
01 Client sends a GET request to `https://my-bucket.s3.us-east-1.amazonaws.com/file.jpg`.
02 S3 endpoint routes the request to the bucket's region and authenticates via IAM or a presigned URL.
03 Object store looks up the object in the bucket's key-value index and retrieves it from replicated storage across >= 3 AZs.
04 Response streams the object data + headers (ETag, Content-Type, version ID) back to the client.
POKE IT YOURSELF
aws s3 ls s3://my-bucket/ — list objects in a bucket
aws s3 cp file.txt s3://my-bucket/ — upload a file to S3
Drill this topic →
~70 sec read