PostgreSQL

PostgreSQL has become the de facto standard for relational databases. For an SMB, it's the database I recommend by default whenever there's data to store: e-commerce sites, business applications, analytical dashboards. Its 30-year robustness combines with rare modernity: native JSON, full-text search, geolocation, and now vector search for AI. Where MySQL/MariaDB remain valid, PostgreSQL almost always offers more headroom at no extra cost.

• →MySQL / MariaDBIf you already have MySQL and migration doesn't bring enough value, valid solution but less modern
• →SQLiteFor very simple applications or prototypes: single file, zero server
• →MongoDBIf your data is strictly documentary without relations, uncommon in SMBs
• →SupabaseManaged PostgreSQL + auth + storage as SaaS: good to start fast, watch long term for cost

1. 01

   Recent LTS version (15 or 16 currently): avoid versions too old for security and performance

2. 02

   Versioned migrations in code (Prisma, Drizzle, Doctrine, Eloquent or native tools)

3. 03

   Automated encrypted backups with Restic or pg_dump + S3, regularly tested

4. 04

   Indexes designed based on actual queries, not by superstition

5. 05

   Basic slow query monitoring, often the most profitable optimization

• PostgreSQL or MySQL?

  PostgreSQL in the vast majority of cases. PostgreSQL has better SQL standard compliance, more modern JSON handling, more native features (full-text, geo, vector). MySQL/MariaDB remain valid for existing or very simple projects. For a new SMB project, I systematically go with PostgreSQL.
• How much does PostgreSQL hosting cost?

  For an SMB, expect €10-40/month self-hosted on a VPS, or €25-100/month on a managed service (OVH, Scaleway, Supabase). Cloud-managed databases (AWS RDS, GCP Cloud SQL) start around €50/month and grow quickly. For most SMBs, self-hosted with automated backups is the best cost/control ratio.
• ORM or raw SQL?

  Depends on the project. For a complex business app (Symfony, Laravel, Next.js), an ORM (Doctrine, Eloquent, Prisma, Drizzle) accelerates development and secures queries. For analytical or performance-critical queries, raw SQL remains preferable. Good frameworks allow mixing both without dogma.
• How to migrate from an old database to PostgreSQL?

  Three steps: structured extraction (often CSV or SQL dump), schema cleanup (inconsistent types, missing constraints), import into PostgreSQL with verifications. For a common MySQL/MariaDB database, this is typically 1-5 days depending on size and cleanliness. For more exotic databases (Access, FoxPro), allow more discovery time.
• And backups?

  The critical issue too often neglected. Daily automated backups at minimum, encrypted at rest (Restic or pg_dump + encrypted S3), retained 30 days in rotation, and regularly tested by real restoration. An untested backup is not a backup. That's the classic trap.