Scale by adding replicas

Throughput grows as the fleet grows — and the 1:10 app-to-worker ratio keeps the app tier from becoming the wall.

4 apps

1 vCPU each

·

40 workers

concurrency 1

93.5req/s

2× fleet

8 apps

1 vCPU each

·

80 workers

concurrency 1

148.9req/s

Measured on GKE · same-region bucket · warm (AP_REUSE_SANDBOX=true)