Storage & Databases

Why not files? How databases actually store data — row vs column orientation, heap files, and the storage engine basics that explain everything above.

Atomicity, consistency, isolation, durability. Transaction isolation levels, distributed transactions, 2PC, and Saga — everything about keeping data correct under concurrency.

Relational model, normalisation, denormalisation tradeoffs, joins, and views. The fundamentals before you decide SQL isn't enough.

Hash indexes, B+ Trees, LSM Trees, and geospatial indexing. The data structures that make reads fast and writes expensive — and when each one wins.

Sync vs async replication, replication lag, failover, and multi-primary. How you keep multiple copies of data consistent without destroying write throughput.

Shard keys, sharding strategies, consistent hashing, cross-shard joins, resharding, and over-sharding. How you split a database that's too big for one node.

Multi-Version Concurrency Control. How databases serve reads without blocking writes by keeping multiple versions of the same row alive simultaneously.

Change Data Capture and the outbox pattern. How you stream database changes to downstream systems without dual-write bugs.

Offset vs cursor pagination. Why offset pagination breaks at scale and how cursor-based pagination solves it without re-scanning the whole table.

The real cost of a database connection and how connection pools amortize it. What happens when the pool is exhausted and how to size it correctly.

Routing reads to replicas and writes to primary. The replication lag problem that follows, and when read/write splitting actually helps vs hurts.

Key-value, document, column-family, search engines, graph, blob storage, NewSQL, OLTP vs OLAP. Every database type, when it wins, and when it doesn't.

A decision framework and cheatsheet for picking the right database given your access patterns, consistency needs, and scale requirements.

Entities, relationships, access patterns, and red flags. How to model data for a real system — with a worked Instagram schema as the example.