E05: How much load the origin can take
- Question
- Where does nginx on a 1 vCPU droplet stop keeping latency low, and how much headroom is that?
- Prediction
- A knee around 3–8k requests/s for HTML, 100–1000× more than a front-page traffic spike would send to the origin. Written before the first run.
- Method
- k6 at fixed arrival rates against a loopback-only listener on the droplet, with vmstat alongside.
Results
Not run yet. The prediction above stands until it is.
What I learned
Not run yet.