Orders-of-magnitude

The full survey of everything you can reach for when you need to scale, on both sides of the wire. API-side levers from horizontal scale and caching to async and backpressure, and client-side levers from the CDN to optimistic UI. Each one gets a line on what it buys and what it costs. This is the map; the API and UI series are the territory.

A walk up the orders of magnitude with one system. At hundreds of requests per second a single box is fine. At thousands, caching and replicas earn their keep. At tens of thousands the sync path cracks and state becomes the enemy. At millions, physics starts dictating the architecture. What breaks first at each rung, and the cheapest thing that fixes it.

The first part of a pillar on scaling. Before you design anything, you have to read the numbers honestly: peak versus sustained, the read/write mix, tail latency, and the fan-out that turns one user action into dozens of downstream calls. The raw request-per-second figure almost always lies.