system-design-primer
simulation 03read storm

Traffic is 95% reads, and every one of them reaches a database that serves eight at a time and takes 70ms each. More servers won't help — the bottleneck is downstream. Get the tail latency down.

pass when
error rate ≤ 3% · p99 ≤ 260ms · spend ≤ $460

The simulation reads best on a wider screen — the topology is drawn left to right.

Ready to run.

served
0
dropped
0
error rate
0.0%
p99
0ms
in flight
0
build
spend$100 / $460

Capacity is never free. Passing under budget is the whole exercise.