Skip to content

Base Architecture

The simplest thing that works#

The base architecture is as simple as it gets:

Client → ID Generator (in-memory counter) → Client

A single ID generator server maintains an in-memory counter starting at 1. Every incoming request reads the current counter value, increments it by 1, and returns it. No database, no coordination, no randomness.

This is 100% unique on a single server — no two requests can ever get the same number because the counter only moves forward.

Base architecture is intentionally naive

The point of the base architecture is to establish the simplest correct system, not the most scalable one. The concurrency problem, the multi-server problem, and the scale problem are all covered in the deep dives.

What breaks next#

Two things break as soon as you try to scale:

  1. Concurrency on a single server — simultaneous requests collide on the counter (the lock problem above)
  2. Multiple servers — two servers with independent counters will both generate ID 1, 2, 3... and collide with each other immediately

Both of these are deep dive problems. The base architecture just establishes the mental model: a counter that only moves forward is the foundation of every ID generation scheme.