What Is Amazon S3? S3 vs EBS vs EFS, Storage Classes and Pricing Explained

By Ravindra Bagale, trainer · LinkedIn · Last updated: 22 October 2026

1. A photo-app story

Priya builds a small photo-sharing app. She puts it on one EC2 server with a 20 GB EBS disk. Users upload photos into /var/www/uploads.

The app becomes popular. In two months the disk is full. She adds a second server behind a load balancer. Now half the photos are on server 1 and half on server 2. Users see broken images.

She needs one place that every server can reach, that never fills up, that can serve a photo over a URL, and that charges only for what is stored. That place is Amazon S3.

2. The limits of EBS for user files

Why do we need this?

EBS is the disk we already know. Before we leave it, we must see exactly where it hurts, so the choice of S3 makes sense.

What is it?

• One Availability Zone. A volume lives in one AZ. If that AZ has trouble, the data is stuck there.
• One instance at a time. A normal volume attaches to one EC2. Only io1/io2 Multi-Attach can join up to 16 Nitro instances in the same AZ, and the app must handle shared writes.
• Fixed size you provision. You choose 20 GB or 500 GB up front.
• You pay for what you provision, not what you use. A 500 GB volume with 50 GB of files is billed for 500 GB.
• Not reachable over HTTP. A browser cannot open a file on EBS. A server must read it and send it.
• Scaling by hand. When it fills, you modify the volume, then run growpart and xfs_growfs.

3. The limits of EFS for user files

Why do we need this?

EFS fixes the "one server" problem: many servers can mount it. So why not use EFS for photos? Because it has its own costs and limits.

What is it?

Amazon EFS (Elastic File System) is a shared network folder (NFS) that grows by itself.

• Linux and NFS only. Windows servers cannot mount EFS (Windows uses FSx).
• Higher price per GB. EFS Standard is about $0.30 per GB-month in us-east-1, more than ten times S3 Standard.
• Inside the VPC. You need a mount target in each AZ and NFS port 2049 open in security groups.
• Network latency. Every read goes over the network, slower than a local EBS disk for small files.
• Not for public web delivery. Like EBS, a browser cannot open an EFS file by URL.

4. S3 vs EBS vs EFS: which one when?

Why do we need this?

Each storage type is best at one job. Picking the right one saves money and work.

What is it?

5. What is Amazon S3? Buckets, objects, keys and prefixes

Why do we need this?

Every S3 command uses these words. If you know them, every later chapter becomes easy.

What is it?

Amazon S3 (Simple Storage Service) stores files as objects inside buckets. Think of a bucket as a big locker room and each object as a parcel with a label.

• No real folders. S3 is flat. users/42/beach.jpg is one key; the slash only looks like a folder.
• Unlimited total storage. One object can be up to about 50 TB (48.8 TiB). AWS raised this from 5 TB in late 2025. The console uploads up to 160 GB, and one PUT up to 5 GB; bigger files use multipart upload.
• Bucket names: 3 to 63 characters, lowercase letters, numbers, dots and hyphens. An account can create 10,000 general purpose buckets by default.
• Strong consistency. Right after you upload or overwrite, every read gets the new data.

6. Durability vs availability

Why do we need this?

AWS uses both words on every storage page. People mix them up and choose the wrong class.

What is it?

• Durability = will my file still exist? S3 is designed for 99.999999999% (11 nines). Store 10 million objects and you may lose one in about 10,000 years.
• Availability = can I read it right now? S3 Standard is designed for 99.99%, about 53 minutes of possible trouble in a year.

Example: a bank locker. Durability says your gold is never lost. Availability says the bank is open when you visit. A closed bank does not mean lost gold.

S3 gets this durability by keeping copies across at least three AZs (except One Zone classes) and checking them all the time.

7. S3 storage classes and prices

Why do we need this?

Not every file is opened every day. Old invoices and last year's logs are rarely read. Paying the top price for them wastes money.

What is it?

A storage class is a price plan for an object. Cheaper classes charge more to read and keep objects for a minimum time. Prices below are approximate for us-east-1 per GB-month; Mumbai is a little higher. Always check the S3 pricing page.

IA and Glacier Instant classes also have a minimum billable object size of 128 KB and a retrieval charge per GB.

Why S3 is economical: a worked example

1. Priya has 1 TB (1,024 GB) of photos. On S3 Standard: 1,024 × $0.023 = about $23.55 per month.
2. The same space as a provisioned gp3 EBS volume: 1,024 × $0.08 = about $81.92, even if half of it is empty.
3. On EFS Standard: 1,024 × $0.30 = about $307.
4. Photos older than a year move to Glacier Deep Archive by a lifecycle rule: 1,024 × $0.00099 = about $1.01.

Three reasons S3 is cheap: you pay only per GB used and per request, you move cold data to cheaper classes (tiering), and lifecycle rules do that move automatically.

8. Upload a file from many stacks

Why do we need this?

Your app talks to S3 through an SDK, not through the console. Every language has one, and they all do the same PutObject call.

What is it?

9. Worked examples, simplest first

1. Copy one file. aws s3 cp notes.txt s3://priya-photos-2026/ → key notes.txt.
2. Copy into a prefix. aws s3 cp beach.jpg s3://priya-photos-2026/users/42/ → key users/42/beach.jpg.
3. Copy a folder. aws s3 cp ./site s3://priya-photos-2026/site/ --recursive uploads every file.
4. Sync only changes. aws s3 sync ./site s3://priya-photos-2026/site/ uploads new or changed files only. Run it twice; the second run uploads nothing.
5. Choose a cheaper class. aws s3 cp old.zip s3://priya-photos-2026/archive/ --storage-class GLACIER_IR.

10. Lab: create a bucket and use it from the console and CLI

Why do we need this?

Reading is not enough. You will create a bucket, upload, list, sync and clean up, and see every result yourself.

How to do it (step by step)

Part A: console

1. Sign in to the AWS console. Top right, choose Region Asia Pacific (Mumbai) ap-south-1. Why: the bucket lives in this Region.
2. Search S3 and open it. Click Create bucket.
3. Bucket type: General purpose. Why: normal buckets; directory buckets are for Express One Zone.
4. Bucket name: yourname-s3-lab-2026 (lowercase, unique). Why: names are global; add your name so it is free.
5. Object Ownership: keep ACLs disabled (recommended). Why: permissions come only from policies, simpler and safer.
6. Block Public Access: keep Block all public access ticked. Why: nothing becomes public by mistake.
7. Bucket Versioning: Disable for this lab. Why: we learn versioning in a later chapter.
8. Default encryption: SSE-S3, Bucket Key enabled. Why: every object is encrypted at rest at no extra cost.
9. Click Create bucket. Expected: a green banner "Successfully created bucket".
10. Open the bucket. Click Create folder, name photos. Why: creates the prefix photos/.
11. Open photos/, click Upload › Add files, pick any small image, click Upload. Expected: "Upload succeeded".
12. Click the object. Copy the Object URL and open it in a new tab. Expected: AccessDenied XML. Why: the bucket is private, which is what we want.
13. Click Open at the top of the object page. Expected: the image opens. Why: the console gives you a short-lived signed link.

Part B: AWS CLI (CloudShell or an EC2 with an S3 role)

1. Open CloudShell from the console top bar. Why: CLI and your login are ready, no keys needed.
2. List buckets:

aws s3 ls
# 2026-10-22 10:30:12 yourname-s3-lab-2026

3. Make two test files and upload one:

mkdir site && echo "<h1>Hello S3</h1>" > site/index.html && echo "v1" > site/about.txt
aws s3 cp site/index.html s3://yourname-s3-lab-2026/site/
# upload: site/index.html to s3://yourname-s3-lab-2026/site/index.html

4. Sync the whole folder:

aws s3 sync site s3://yourname-s3-lab-2026/site/
# upload: site/about.txt to s3://yourname-s3-lab-2026/site/about.txt
aws s3 sync site s3://yourname-s3-lab-2026/site/
# (no output: nothing changed)

5. Verify with ls (recursive, human sizes):

aws s3 ls s3://yourname-s3-lab-2026 --recursive --human-readable --summarize
# 2026-10-22 10:41:03   18 Bytes site/index.html
# 2026-10-22 10:41:20    3 Bytes site/about.txt
# Total Objects: 3

6. See the storage class and encryption of one object:

aws s3api head-object --bucket yourname-s3-lab-2026 --key site/index.html
# "ServerSideEncryption": "AES256", "ContentLength": 18 ...

7. Download it back: aws s3 cp s3://yourname-s3-lab-2026/site/index.html check.html && cat check.html. Expected: <h1>Hello S3</h1>.

Cleanup

1. Delete all objects: aws s3 rm s3://yourname-s3-lab-2026 --recursive.
2. Delete the bucket: aws s3 rb s3://yourname-s3-lab-2026. Expected: remove_bucket: yourname-s3-lab-2026. Why: a bucket must be empty before it can be deleted.

11. Interview questions

Revision, comparison and interview cards

Revise this lesson in a few minutes. Save the cards on your phone or download them as PNG files. The QR code on each card opens this lesson.

More cards: Ravi Sir's Cards.