Open the simulator
Free, open source, and it runs entirely in your browser.
Why the numbers can be trusted
The engine underneath is a discrete-event simulator, not a formula that approximates one. Requests are individual entities that arrive, queue, get served and time out. When a page here says a design loses every request at 113 a second, that is 113 simulated requests a second actually failing. Vendor pricing is cited from published rates with the date each was read. A teaching tool that shows a plausible number is worse than one that shows no number, because you cannot tell the difference and will carry the wrong intuition into real work. Every table in these examples is measured at build time by running the design through the same engine you use in the app, at a fixed seed. They cannot drift from the tool, because nobody types them.Where to start
Retry storm
A short timeout with retries in front of a small database. Healthy at 90
requests a second, serving nothing at 113. The clearest cliff in the set.
Cache aside
How much traffic never reaches your database, and what happens the moment it
does.
Discord
Millions of sockets and fan-out per message. The gateway runs out of
connections, not requests.