Recipes
Concrete, reusable techniques for common scaling problems — the practical "how" behind the patterns covered elsewhere in this repo.
The other sections of this repo explain the big building blocks — caching, messaging, load balancing. Recipes are smaller and more targeted: each one is a self-contained technique you can reach for when a specific problem shows up, usually a place where the naive implementation quietly stops scaling.
A recipe answers a narrow question: "I'm doing X the obvious way and it's melting under load — what's the standard fix?" The answer almost always involves trading a little immediacy or simplicity for a large win in throughput, cost, or stability.
What's in this section
- Batching & Aggregation — stop doing one database write per event. Buffer high-frequency, low-value writes (likes, views, metrics) and flush them as a single batched or aggregated operation.
More recipes will be added over time. Each one is written to stand on its own, so you can jump straight to the one you need.
Invalidation and Consistency
Keeping the cached copy honest — invalidation strategies, the consistency you actually get, and surviving stampedes, penetration, and avalanches.
Batching & Aggregation
Stop running one write per event. Buffer high-frequency, low-value writes — likes, views, metrics — and flush them as a single batched insert or an aggregated counter update.