AWS Builder Center

Intro to S3

I've been using S3 for close to eight years now, and I still find myself explaining it to people the same way: it's a place to put files that you never have to think about again. No servers to patch, no disks to worry about filling up, no midnight page because a drive failed.

Senior Software Engineer
S3 confuses people who come from a "server with a hard drive" mental model, so here's the short version: it's object storage. You put a file in, you get a URL-ish key back, and you stop thinking about disks entirely. That's the whole pitch, and it holds up.
An object is just data plus a key (its path-like name) plus some metadata, living in a bucket. Up to 5TB per object. There's no real folder structure — the "folders" in the console are cosmetic, built from keys that happen to contain slashes. This trips people up constantly when they try to "move a folder" and realize there's no such operation, just a batch rename.
Storage classes are where the money actually gets saved or wasted. Standard makes sense for stuff you touch often. Everything else is a spectrum of cheaper-but-slower-to-retrieve. I moved a client's old backup bucket to Glacier Deep Archive and their monthly bill dropped by roughly 70%. Worth knowing before you do it: retrieval from Deep Archive can take the better part of a day, so it's for things you're confident you won't need urgently.
Turn on versioning. I say this to nearly everyone, and nearly everyone ignores it until after they've overwritten or deleted something they needed. It keeps every prior version of an object instead of clobbering it on write. Small extra storage cost, real insurance.
On security — buckets are private by default now, and Block Public Access is on unless someone deliberately turns it off. That's a real improvement over a few years ago, when "leaked S3 bucket" was basically a recurring headline. Even so, I still check bucket policies by hand on anything holding customer data. Defaults are good; they're not a substitute for looking.
As a rule of thumb: if the data isn't going into a relational database, it probably belongs in S3 — static assets, backups, logs, training data, video, whatever. You pay for storage plus requests plus transfer, and that's usually cheap enough that it's not worth second-guessing early on.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article