Database Options for OJS

Open Journal Systems uses a relational database to store all journal content, submissions, users, and workflow state. Choosing the right database engine — and understanding the broader database landscape — helps site administrators make informed decisions about performance, scalability, and long-term maintenance.

1. Databases Officially Supported by OJS

OJS supports two RDBMS engines out of the box. The choice is made in config.inc.php before running the installer.

1.1 MySQL 8.x

Property	Value
Type	RDBMS
Licence	GPL (Community) / Commercial (Enterprise)
OJS driver	`mysqli`
Character set	`utf8mb4` (required for emoji / full Unicode)
PKP default	Yes — most community documentation assumes MySQL

MySQL is the most widely used database for OJS installations. It is included in every major Linux distribution's package repositories and supported by all control panels (cPanel, Plesk, HestiaCP, CyberPanel).

Recommended config.inc.php for MySQL:

driver = mysqli
host = localhost
username = ojsuser
password = your_password
name = ojs_production

Minimal my.cnf tuning for OJS:

[mysqld]
character-set-server = utf8mb4
collation-server = utf8mb4_unicode_ci
innodb_buffer_pool_size = 512M   ; Set to ~50-70% of available RAM
innodb_log_file_size = 128M
max_allowed_packet = 128M
query_cache_type = 0             ; Disable query cache (deprecated in MySQL 8)

Source: PKP Admin Guide – System Requirements

1.2 PostgreSQL

Property	Value
Type	RDBMS — Advanced Features
Licence	PostgreSQL Licence (permissive open source)
OJS driver	`postgres9`
PKP support	Fully supported; some PKP developers prefer it
Strengths	ACID compliance, JSON support, advanced indexing, full-text search

PostgreSQL is the preferred choice for larger installations, multi-journal deployments, or administrators who need advanced query features. It is also better suited for read-heavy workloads with complex analytics (e.g., COUNTER statistics).

Recommended config.inc.php for PostgreSQL:

driver = postgres9
host = localhost
username = ojsuser
password = your_password
name = ojs_production

Database creation:

CREATE USER ojsuser WITH PASSWORD 'your_password';
CREATE DATABASE ojs_production
  OWNER ojsuser
  ENCODING 'UTF8'
  LC_COLLATE 'en_US.UTF-8'
  LC_CTYPE 'en_US.UTF-8';
GRANT ALL PRIVILEGES ON DATABASE ojs_production TO ojsuser;

postgresql.conf performance tuning:

shared_buffers = 512MB           # ~25% of RAM
work_mem = 32MB
maintenance_work_mem = 128MB
effective_cache_size = 1GB
random_page_cost = 1.1           # For SSD storage
checkpoint_completion_target = 0.9

Source: PKP Community Forum – PostgreSQL discussions

2. MySQL-Compatible Databases (Drop-in Replacements)

Several databases are wire-compatible with the MySQL protocol and can be used with OJS's mysqli driver without code changes.

2.1 MariaDB

Property	Value
Type	RDBMS — MySQL compatible
Licence	GPL
OJS support	Works with `mysqli` driver; widely used in the community
Default in	Ubuntu (when installed via `apt install mariadb-server`)

MariaDB is a community-developed fork of MySQL, created by MySQL's original developer (Monty Widenius) after Oracle acquired MySQL. It is a drop-in replacement for MySQL and the default database on many Linux distributions and control panels.

Key differences from MySQL 8:

No authentication change to caching_sha2_password (uses mysql_native_password by default — important for some PHP versions)
Slightly different query optimizer (sometimes faster for reporting)
Galera Cluster for multi-master replication (useful for high-availability OJS)
CONNECT storage engine for CSV/JSON data

MariaDB is recommended over MySQL for: most OJS deployments on Ubuntu/Debian, users of HestiaCP or CyberPanel, and anyone wanting simpler apt-based management.

Source: MariaDB Documentation

2.2 Percona Server for MySQL

Property	Value
Type	RDBMS — MySQL-compatible, enhanced
Licence	GPL
OJS driver	`mysqli` (compatible)
Strengths	XtraDB storage engine, superior InnoDB performance, audit logging, Percona Toolkit

Percona Server is a drop-in replacement for MySQL with enhanced performance, more detailed monitoring, and enterprise-grade features at no cost. It is particularly suitable for high-volume OJS multi-journal installations where InnoDB performance is critical.

When to use Percona:

You have 10+ active journals on a single installation
You need detailed MySQL performance metrics (Percona Monitoring and Management — PMM)
You require audit logging for compliance (HIPAA, GDPR)
You want automatic schema change without downtime (Percona Online Schema Change)

Source: Percona Server Documentation

2.3 TiDB

Property	Value
Type	Distributed RDBMS — MySQL compatible
Licence	Apache 2.0
OJS driver	`mysqli` (TiDB is MySQL protocol compatible)
Architecture	Distributed SQL — scales horizontally across nodes
Use case	Very large scale, cloud-native, multi-region

TiDB is a distributed SQL database that is MySQL-wire-compatible. It is designed for very large data volumes (billions of rows) and cloud-native deployments with horizontal scaling. For typical OJS deployments (even large multi-journal setups), TiDB is over-engineered. However, it could be relevant for a national repository running hundreds of journals with millions of submissions.

OJS + TiDB considerations:

TiDB does not support all MySQL features (e.g., some stored procedures, FULLTEXT indexes)
OJS has not been officially tested with TiDB; test thoroughly before production use
Suitable only for very large institutional repositories, not for individual journal hosting

Source: TiDB Documentation

2.4 CockroachDB

Property	Value
Type	Distributed RDBMS — PostgreSQL compatible
Licence	BSL (core) / Enterprise
OJS driver	`postgres9` (CockroachDB speaks PostgreSQL wire protocol)
Architecture	Distributed; automatic replication, geo-partitioning

CockroachDB is a distributed PostgreSQL-compatible database designed for global applications with automatic failover. Like TiDB, it is significantly over-engineered for most OJS deployments but could be relevant for very large-scale national or international platforms.

OJS + CockroachDB considerations:

PostgreSQL wire protocol means OJS's postgres9 driver should work
CockroachDB does not support all PostgreSQL extensions (e.g., specific aggregate functions)
Requires careful testing; not officially supported by PKP

3. Database Landscape: NoSQL and Other Systems

The following databases are commonly encountered in web infrastructure discussions. While none of them are supported as OJS primary databases, understanding them helps administrators who run OJS alongside other applications or services.

3.1 SQLite

Property	Value
Type	RDBMS — Lightweight, embedded
Licence	Public Domain
OJS support	❌ Not supported for OJS
Use cases	OJS local development / testing only, embedded applications

SQLite stores the entire database in a single file. It is not suitable for production OJS because:

No concurrent write support (single-writer model)
Not designed for multi-process web server access
No network access (can't be used with separate app/database servers)

SQLite can be useful for local development if you run OJS via Docker and want a zero-config database for quick testing, but PKP does not officially support it.

3.2 MongoDB

Property	Value
Type	NoSQL — Document storage
Licence	SSPL (Server Side Public Licence)
OJS support	❌ Not supported
Strengths	Schema flexibility, horizontal scaling, JSON documents

MongoDB stores data as BSON (binary JSON) documents. It is widely used for content management applications, but OJS's relational schema (foreign keys between submissions, users, journals, review rounds) is fundamentally relational and does not map well to a document model. OJS cannot use MongoDB as its primary database.

Where MongoDB is relevant to an OJS ecosystem:

Storing analytics/telemetry data alongside OJS
An API layer that caches OJS content for high-read scenarios
Companion applications (e.g., author profile aggregation)

3.3 Redis

Property	Value
Type	NoSQL — In-memory key-value store
Licence	RSALv2 / SSPL (v7.4+), earlier versions BSD
OJS integration	✅ Supported as session cache and object cache

Redis is NOT an OJS primary database. However, it is directly relevant to high-performance OJS deployments as:

PHP Session Storage: OJS can use Redis to store PHP sessions, enabling sticky-session-free horizontal scaling across multiple app servers.
Object Cache: With an object caching layer, repetitive database queries (e.g., journal configuration, navigation menus) are served from RAM.

Enable Redis for PHP sessions:

; /etc/php/8.2/fpm/pool.d/ojs.conf
php_value[session.save_handler] = redis
php_value[session.save_path] = "tcp://127.0.0.1:6379"

Source: PKP Forum – Redis session caching

3.4 Apache Cassandra

Property	Value
Type	NoSQL — Wide-column distributed store
Licence	Apache 2.0
OJS support	❌ Not supported
Strengths	Linear write scalability, no single point of failure, multi-datacenter

Cassandra excels at time-series data, IoT, and large-scale write-heavy workloads. It is not suitable for OJS because OJS requires complex JOIN queries and relational consistency that Cassandra does not support.

3.5 Couchbase Server

Property	Value
Type	NoSQL — Document storage + Key-value
Licence	BSL / Enterprise
OJS support	❌ Not supported
Notable feature	Mobile sync (Couchbase Lite); N1QL query language

Couchbase is a document database designed for mobile-first applications and offline sync. Not applicable to OJS's server-side relational data model.

3.6 ArangoDB

Property	Value
Type	Multi-model — Graph, document, key-value
Licence	BSL (core)
OJS support	❌ Not supported
Strengths	Single engine for graph, document, and key-value queries (AQL query language)

ArangoDB's multi-model nature makes it attractive for applications requiring graph traversal (e.g., citation networks, co-authorship graphs). While not an OJS database, it could be used in a companion service that analyses citation relationships across journal content.

3.7 RethinkDB

Property	Value
Type	NoSQL — Document storage, real-time push
Licence	Apache 2.0
OJS support	❌ Not supported
Notable feature	Real-time changefeeds — database pushes updates to clients without polling

RethinkDB is well-suited for real-time collaboration applications (collaborative editing, live dashboards). Its development has slowed significantly in recent years. Not applicable to OJS.

3.8 Cloudant (IBM)

Property	Value
Type	NoSQL — Distributed document store (CouchDB-based)
Licence	Commercial (IBM Cloud service)
OJS support	❌ Not supported
Notable feature	Managed CouchDB; HTTP API; geo-distributed replication

IBM Cloudant is a managed CouchDB-as-a-service. Not applicable to OJS's relational model.

3.9 OrientDB

Property	Value
Type	Multi-model — Graph and document
Licence	Apache 2.0 (Community) / Commercial
OJS support	❌ Not supported
Notable feature	Supports SQL-like query syntax over graph data

OrientDB combines graph database capabilities with document storage. Suitable for citation networks or knowledge graphs, not for OJS operational data.

3.10 MarkLogic

Property	Value
Type	Multi-model — Enterprise document/search
Licence	Commercial
OJS support	❌ Not supported
Notable feature	Built-in semantic search, ACID transactions at scale, XQuery/JavaScript APIs

MarkLogic is used by large publishers (Elsevier, Springer Nature) for managing massive document corpora. Its XML-native storage and semantic capabilities make it suitable for full-text archives of scholarly literature — but as a document management layer, not as an OJS backend.

3.11 Microsoft SQL Server

Property	Value
Type	RDBMS — Enterprise
Licence	Commercial (Microsoft)
OJS support	❌ Not officially supported
Available on	Windows and Linux

Microsoft SQL Server is a mature enterprise RDBMS. OJS does not have an MSSQL driver. Running OJS on SQL Server would require significant porting effort. Not recommended; use MySQL/MariaDB or PostgreSQL.

3.12 Oracle Database

Property	Value
Type	RDBMS — Enterprise
Licence	Commercial (Oracle)
OJS support	❌ Not supported

Oracle Database is widely used in large enterprises and universities. OJS does not include an Oracle driver. Not applicable to standard OJS deployments.

4. Database Comparison Summary

Database	Type	OJS Support	Best For OJS Context
MySQL 8.x	RDBMS	✅ Official	Most deployments; excellent community support
PostgreSQL	RDBMS	✅ Official	Large scale, advanced queries, statistics
MariaDB	RDBMS (MySQL-compat)	✅ Community	Ubuntu/Debian VPS; drop-in MySQL replacement
Percona Server	RDBMS (MySQL-compat)	✅ Community	High-volume multi-journal; enterprise monitoring
TiDB	Distributed RDBMS (MySQL-compat)	⚠️ Experimental	Very large repositories only
CockroachDB	Distributed RDBMS (PG-compat)	⚠️ Experimental	Geo-distributed, high availability
SQLite	RDBMS (embedded)	❌ No	Dev/testing only
MongoDB	NoSQL document	❌ No	Companion analytics services
Redis	NoSQL in-memory	✅ As cache only	PHP session store, object cache layer
Cassandra	NoSQL wide-column	❌ No	External analytics at scale
Couchbase	NoSQL document	❌ No	Mobile sync apps
ArangoDB	Multi-model (graph)	❌ No	Citation graph analysis tools
RethinkDB	NoSQL real-time	❌ No	Real-time dashboards
Cloudant	NoSQL (CouchDB)	❌ No	Managed document store
OrientDB	Multi-model (graph)	❌ No	Graph-based citation tools
MarkLogic	Multi-model enterprise	❌ No	Large publisher document archives
SQL Server	RDBMS enterprise	❌ No	Not applicable
Oracle DB	RDBMS enterprise	❌ No	Not applicable

5. Choosing the Right Database for OJS

Your Situation	Recommended Database
New OJS installation on Ubuntu/Debian VPS	MariaDB (default apt package, drop-in MySQL)
New OJS installation on AlmaLinux/Rocky/cPanel	MySQL 8.x (standard for RHEL-based systems)
Large multi-journal platform (10+ journals)	PostgreSQL or Percona Server
Need advanced statistics and reporting	PostgreSQL (superior query planner and COUNTER stats)
High-availability, zero-downtime upgrades	Percona XtraDB Cluster or MariaDB Galera Cluster
Cloud-native containerised deployment	MySQL 8.x or PostgreSQL in managed cloud DB services
Quick local development/testing	MariaDB in Docker Compose (official `pkp/docker-ojs` image)

6. Backup Essentials

For whichever database you choose, establish automated daily backups. See the OJS Server Infrastructure article for complete backup scripts covering both MySQL/MariaDB and PostgreSQL.

Quick reference:

# MySQL / MariaDB / Percona backup
mysqldump --single-transaction --routines --events \
  --defaults-file=/etc/mysql/backup.cnf ojs_production \
  | gzip > /backup/ojs_$(date +%Y%m%d).sql.gz

# PostgreSQL backup
PGPASSFILE=/etc/postgresql/.pgpass \
  pg_dump -U ojsuser -F c ojs_production \
  > /backup/ojs_$(date +%Y%m%d).pgdump

References

PKP Admin Guide – System Requirements
https://docs.pkp.sfu.ca/admin-guide/en/
MariaDB Documentation
https://mariadb.com/kb/en/
PostgreSQL Documentation
https://www.postgresql.org/docs/
MySQL 8 Documentation
https://dev.mysql.com/doc/refman/8.0/en/
Percona Server Documentation
https://docs.percona.com/percona-server/latest/
TiDB Documentation
https://docs.pingcap.com/tidb/stable
CockroachDB Documentation
https://www.cockroachlabs.com/docs/
Redis Documentation
https://redis.io/docs/
PKP docker-ojs (GitHub) — Official Docker images with MySQL and PostgreSQL support.
https://github.com/pkp/docker-ojs
PKP Community Forum – Database discussions
https://forum.pkp.sfu.ca/

1. Databases Officially Supported by OJS​

1.1 MySQL 8.x​

1.2 PostgreSQL​

2. MySQL-Compatible Databases (Drop-in Replacements)​

2.1 MariaDB​

2.2 Percona Server for MySQL​

2.3 TiDB​

2.4 CockroachDB​

3. Database Landscape: NoSQL and Other Systems​

3.1 SQLite​

3.2 MongoDB​

3.3 Redis​

3.4 Apache Cassandra​

3.5 Couchbase Server​

3.6 ArangoDB​

3.7 RethinkDB​

3.8 Cloudant (IBM)​

3.9 OrientDB​

3.10 MarkLogic​

3.11 Microsoft SQL Server​

3.12 Oracle Database​

4. Database Comparison Summary​

5. Choosing the Right Database for OJS​

6. Backup Essentials​

References​