What sets Ceph apart from simple object storage like MinIO?

Ceph is unified distributed storage: it exposes object (S3), block, and filesystem over the same cluster, replicates and self-heals, and scales to petabytes. That power also brings operational complexity.

In which scenario does Ceph make sense?

Ceph shines when you need massive, fault-tolerant, multi-protocol storage on your own infrastructure — with people to operate it. For learning or local use, MinIO is far simpler.

If you just want to understand object storage, what's best?

The concept (buckets, objects, the S3 API) you learn in minutes with MinIO. Ceph is the next level when the problem is scale and operations, not learning.

← The catalogue

Nº09 · Storage

Ceph

Distributed storage at production scale: objects, blocks and files.

Storage—Intermediate—Data Engineer

What is it?

Ceph is an open-source distributed storage platform built for production scale. Over a single cluster it offers three interfaces: object (S3-compatible), block (disks for machines/containers), and filesystem. It replicates data across nodes, self-heals on disk failures, and scales to petabytes.

It is the choice when storage itself is the serious problem: a private cloud, large on-premise infrastructure, persistent storage for Kubernetes. That power comes at a cost: operating it is complex and usually requires a dedicated team.

What is it for?

Private storage cloud. A single platform for object, block, and files in your own datacenter.
On-premise lake at scale. Via its RADOS Gateway it exposes the S3 API, so your lake (Parquet, Iceberg) can live on Ceph just like on S3.
Persistent storage for Kubernetes. With Rook, Ceph provides volumes for containerized workloads.

When to use it / when not

Use it when the challenge is scale and operations: petabytes, fault tolerance, multi-protocol, your own infrastructure with people to run it. It is the "serious" on-premise storage option.

Think twice —and it almost always will be, for a beginner— when:

You're learning the object storage concept: MinIO gives it to you in a minute, S3-compatible, with no operational complexity.
Your scale is modest or you work locally/single-node: Ceph's overhead isn't justified.
You're in the cloud with managed S3/GCS/Azure: you already have storage at scale without operating it yourself.

Get started in 1 minute

Let's be honest: Ceph does not stand up in a minute — it is a cluster system you deploy and operate carefully. Two paths depending on your goal:

To understand the concept of object storage (buckets, S3 API), start with MinIO: one container and you're set.
For a real Ceph trial, the official path is cephadm on a dedicated host (not your work laptop):

# On a dedicated test host (Linux), not your main machine:
curl -sLO https://download.ceph.com/rpm-18.2.0/el9/noarch/cephadm
sudo python3 cephadm bootstrap --mon-ip <HOST-IP>
# Brings up a minimal cluster + the web dashboard; from there you add disks (OSDs).

The full deployment and operations guide is in the official documentation.

Quick trivia — test what you just read.