What is a Database?

The simplest explanation

A spreadsheet is a database. It has rows (records) and columns (fields). A proper database takes this idea and makes it:

• Much larger, billions of rows, terabytes of data
• Much faster, finds any record in milliseconds using indexes
• Concurrent, handles thousands of reads and writes simultaneously without corruption
• Reliable, survives crashes without losing data (ACID properties)
• Queryable, you can ask complex questions across multiple tables at once

When you create an account on a website, a new row is added to a users table. When you log in, the system finds your row by email and checks your password hash. When you update your profile, it modifies that row. This read/write cycle happens billions of times per second across all the apps you use.

SQL: the language of relational databases

Most databases are queried using SQL (Structured Query Language), pronounced “sequel” or “S-Q-L”. SQL has been the standard since the 1970s and is still essential. More: SQL, MDN Glossary

The four core operations (CRUD):

-- Create: add a new record
INSERT INTO users (name, email, city)
VALUES ('Linda', 'linda@example.com', 'Vienna');

-- Read: retrieve records
SELECT name, email FROM users WHERE city = 'Vienna';

-- Update: modify an existing record
UPDATE users SET city = 'Berlin' WHERE email = 'linda@example.com';

-- Delete: remove a record
DELETE FROM users WHERE email = 'linda@example.com';

More complex queries join multiple tables:

-- Find all orders for users in Vienna
SELECT users.name, orders.total, orders.created_at
FROM orders
JOIN users ON orders.user_id = users.id
WHERE users.city = 'Vienna'
ORDER BY orders.created_at DESC;

SQL reads almost like English. You do not need to master it to be effective, but reading a query and understanding what it does is a valuable skill. Try it interactively: SQLBolt

Relational vs non-relational databases

Relational databases (SQL) store data in related tables. A users table and an orders table are linked, each order has a user_id that references a specific user. This relationship is enforced by the database (foreign key constraint). Examples: PostgreSQL , MySQL, SQLite, MariaDB.

Non-relational databases (NoSQL) store data in other formats:

• Document stores (MongoDB, Firestore): JSON-like documents, flexible schema, good for hierarchical data
• Key-value stores (Redis, DynamoDB in simple mode): ultra-fast lookup by key, excellent for caching and sessions
• Column stores (Cassandra, BigQuery): optimised for reading large datasets, used in analytics
• Graph databases (Neo4j): store relationships as first-class data, used for social networks and recommendation engines
• Vector databases (Pinecone, pgvector, Weaviate): store and query high-dimensional embeddings for AI applications

For most applications, start with PostgreSQL, the most capable, well-supported, and feature-rich open-source relational database. Add Redis as a cache if you need performance. Only reach for specialised databases when you have a specific need they solve. More: PostgreSQL vs MongoDB, DigitalOcean

ACID properties, why databases are reliable

Databases guarantee ACID for transactions, groups of operations that must all succeed or all fail:

• Atomicity: a transaction either completes entirely or not at all. If a payment deducts from your balance but the order record fails to create, the deduction is rolled back.
• Consistency: the database always goes from one valid state to another. Rules (constraints, foreign keys) are always enforced.
• Isolation: concurrent transactions do not interfere with each other. Two people booking the last seat see correct results.
• Durability: once committed, data survives crashes. It is written to disk.

These properties are what make databases trustworthy for financial transactions, medical records, and anything else where data integrity is critical.

Indexes: how queries stay fast

A database with millions of rows needs a way to find records without reading every row. That is what indexes do.

An index on the email column of users lets the database find any user by email in O(log n) time, milliseconds even with 100 million rows, instead of scanning every row linearly. More: Use The Index, Luke is the best free resource on database indexing.

The trade-off: indexes speed up reads but slow down writes (the index must also be updated). Over-indexing hurts write performance. Index the columns you actually query on.

Migrations: changing your schema safely

When your app needs a new column or a new table, you write a migration, a script that modifies the database schema. Migrations:

• Are tracked in version control alongside your code
• Run in order (migration_001_add_users, migration_002_add_orders)
• Can be rolled back if something goes wrong

Every serious ORM has a migration system. In production, migrations need care: adding a column to a 100-million-row table can lock the table for minutes.

Vector databases and AI

Modern AI applications use embeddings, numerical representations of text, images, or audio as high-dimensional vectors. Semantic search, recommendation systems, and RAG (Retrieval-Augmented Generation) all require storing and querying these vectors.

Vector databases efficiently find the most similar vectors to a query:

• Pinecone , managed vector database
• pgvector , PostgreSQL extension for vector storage (same database, no extra service)
• Weaviate , open-source vector database with hybrid search
• Chroma , simple, open-source, good for prototyping

For most AI applications, pgvector in Supabase lets you do vector search in your existing PostgreSQL database without additional infrastructure.

Managed databases: skip the administration

Running your own database server involves backups, updates, replication, and security patches. Cloud providers offer managed databases where they handle all of this:

For a prototype: Supabase or Neon. Both are free, PostgreSQL-based, and work with every major framework.

Further reading

• SQLBolt, interactive SQL lessons , the best free way to actually learn SQL by doing
• PostgreSQL Tutorial , from fundamentals to advanced queries
• Use The Index, Luke , database indexing explained without jargon
• Supabase documentation , excellent docs for getting started with PostgreSQL
• Database Design for Beginners, ByteByteGo , video explanations with clear diagrams
• Prisma (ORM) documentation , if you use TypeScript/JavaScript, Prisma is the most popular ORM

What’s next

Next: What is AI? , what artificial intelligence actually is, how large language models work, and how they fit into everything you have just learned.