What is the best self-hosted databases?

Q: What is the best self-hosted databases?

PostgreSQL unless you have a specific reason otherwise. It runs everything, has the strongest ecosystem, and every hosting question you will ever ask already has an answer for it. Reach for the others when your data has a shape PostgreSQL handles badly.

Best Self-Hosted Databases

This is the most consequential choice on the site and the one people agonise over the least, because most self-hosters inherit whichever database their application requires.

Updated 2026-10-03 · 7 options compared

The short answer

PostgreSQL unless you have a specific reason otherwise. It runs everything, has the strongest ecosystem, and every hosting question you will ever ask already has an answer for it. Reach for the others when your data has a shape PostgreSQL handles badly.

This is the most consequential choice on the site and the one people agonise over the least, because most self-hosters inherit whichever database their application requires. That is fine. What follows is for the cases where you are choosing — a new project, or a migration — and it is organised around data shape rather than benchmarks, because the honest difference in real self-hosted workloads is usually operational, not raw performance.

7 options at a glance

Comparison of the 7 self-hosted options ranked on this page
#	Tool	License	Setup	Best for
1	PostgreSQL	PostgreSQL	Intermediate	Nearly everything relational, and the safe default for a new project.
2	MariaDB	GPL-2.0	Intermediate	PHP applications and workloads built around the MySQL dialect.
3	Valkey	BSD-3-Clause	Intermediate	Caching, queues and session state that can be rebuilt.
4	MongoDB	SSPL	Intermediate	Flexible or rapidly evolving document schemas.
5	ClickHouse	Apache-2.0	Advanced	Analytics and aggregation over very large datasets.
6	Apache CouchDB	Apache-2.0	Intermediate	Offline-first applications that must sync on reconnect.
7	Garage	AGPL-3.0	Advanced	Lightweight geo-distributed S3 storage across a few nodes.

Which one should you pick?

Start with the shape of the data. Relational with integrity constraints: PostgreSQL. Relational and inherited from a decade-old PHP application: MariaDB. Key-value with high read volume and the ability to lose everything on restart: Valkey. Documents with flexible schemas: MongoDB, or FerretDB if you want that API over PostgreSQL storage. Analytical queries over billions of rows: ClickHouse. Flat objects at scale: MinIO, Garage or SeaweedFS. The most common mistake is not picking the wrong engine — it is running a database with no tested backup, which is a data-loss incident on a timer.

What actually decides it

Every database needs an automated backup from day one. Not a replication setup, not a snapshot someone takes manually. A scheduled dump or physical backup, to storage elsewhere, with a restore that has been tested at least once.
Replication is not backup. A replica faithfully reproduces a `DELETE FROM`. It protects against hardware failure and nothing else, which is one of the four ways databases die.
Put the data volume somewhere you can measure. Databases write constantly and fill disks in the middle of the night. Alert on free space before it is a crisis, not after.
Tune one thing and leave the rest alone. For most self-hosted workloads, three settings matter — shared buffers, work mem, and max connections. Changing random parameters found in a forum post is how people make a working database slower.
Do not expose a database port to the internet. Not with authentication, not behind a firewall rule you meant to tighten later. Put it on a private network, or reach it through your VPN.

All 7 options, in order

PostgreSQL

Best for: Nearly everything relational, and the safe default for a new project.

The default answer because it is genuinely that good: rigorous about correctness, feature-rich far beyond what most users touch — JSON, full-text search, geospatial, extensions — and backed by the deepest pool of documentation and community answers of any database. Almost every application you will self-host supports it, and starting here means never migrating for a bad reason.

Databases & Storage PostgreSQL Intermediate ★ 22,262

Replaces: MySQL, MariaDB, Amazon RDS

Full review of PostgreSQL →

MariaDB

Best for: PHP applications and workloads built around the MySQL dialect.

The community continuation of MySQL, and the pragmatic answer for the enormous installed base of PHP applications built against that dialect. It is fast, well understood, and every hosting provider supports it. Choose it when something requires it rather than on its own merits — but when that requirement exists, it is the right choice and not a compromise.

Databases & Storage GPL-2.0 Intermediate ★ 8,308

Replaces: MySQL, PostgreSQL, Amazon Aurora

Full review of MariaDB →

Valkey

Best for: Caching, queues and session state that can be rebuilt.

Under its current name Valkey, and the answer for caches, queues, session stores and rate limiters — the fast, ephemeral layer almost every application wants. The critical thing to understand is that it is designed around data that may be lost: treat it as a cache with a rebuild path, not as your source of truth, and the failure modes stop being alarming.

Databases & Storage BSD-3-Clause Intermediate ★ 27,352

Replaces: Redis Enterprise, Amazon ElastiCache, Memcached

Full review of Valkey →

MongoDB

Best for: Flexible or rapidly evolving document schemas.

Document storage with a flexible schema, which is genuinely the right model when your data is a varied collection of records that would require constant migrations in a relational schema. The licence is a real consideration for commercial use, and FerretDB on this list offers the API without it.

Databases & Storage SSPL Intermediate ★ 28,617

Replaces: Amazon DocumentDB, Firebase Firestore, CouchDB

Full review of MongoDB →

ClickHouse

Best for: Analytics and aggregation over very large datasets.

A columnar engine built for analytical queries — aggregating and slicing enormous datasets — where it is orders of magnitude faster than a row-store. It is not an OLTP database and should not be used as one. Deploy it when you have real analytical volume, because below that threshold the operational cost is not repaid.

Databases & Storage Apache-2.0 Advanced ★ 50,191

Replaces: Google BigQuery, Amazon Redshift, Snowflake

Full review of ClickHouse →

Apache CouchDB

Best for: Offline-first applications that must sync on reconnect.

The database whose headline feature is synchronisation: offline-first clients that reconcile with a server, with a documented conflict model. That makes it the right foundation for an application designed to work offline, a niche where the relational options are considerably more work.

Databases & Storage Apache-2.0 Intermediate ★ 6,967

Replaces: Firebase Firestore, MongoDB, PouchDB Cloud

Full review of Apache CouchDB →

Garage

Best for: Lightweight geo-distributed S3 storage across a few nodes.

S3-compatible object storage that is dramatically lighter to operate than the alternatives, and designed to spread across machines that need not be in the same building. For a home lab with a couple of nodes holding backups and media, it is the right trade: a little less maturity for a lot less machinery.

Databases & Storage AGPL-3.0 Advanced ★ 4,625

Replaces: AWS S3, MinIO, Ceph

Full review of Garage →

How these compare to what you are using now

A managed database

Costs more and hands over a hard dependency. Worth it when you cannot be the one on call, and a lot of self-hosters simply accept being on call.

Best Self-Hosted Databases

7 options at a glance

Which one should you pick?

What actually decides it

All 7 options, in order

PostgreSQL

MariaDB

Valkey

MongoDB

ClickHouse

Apache CouchDB

Garage

How these compare to what you are using now

A managed database

Other rankings

Best Self-Hosted Media Servers

Best Self-Hosted Photo Apps

Best Self-Hosted Note-Taking Apps

Best Self-Hosted Cloud Storage

Best Self-Hosted Backup Tools

Best Self-Hosted Git Servers