One big server vs three small servers
One big server has no spare when it fails. Three small servers share the load, so the other two carry it when one fails. At 3,600 requests/s, one big server reaches only 67 percent availability, and three small servers stay above 99 percent for less money.
280 req/s · CPU 70% · mean 22 ms
Two rows · 3 of 3 replicas up · load served
Case 1
The API serves 3,600 requests/s. One server fails at 0:40. Availability must stay above 95 percent and errors below 5 percent.
| Measure | Goal | One big server | Three small servers |
|---|---|---|---|
| p99 latency | under 250 ms | 24 ms Pass | 69 ms Pass |
| Availability | at least 95% | 66.7% Fail | 99.944% Pass |
| Error rate | under 5% | 33.3% Fail | 0.056% Pass |
| Monthly cost | 32 USD/mo | 14 USD/mo |