Workload t = 0.0s

Counts requests only

Admits while requests-per-minute has room. Never looks at tokens.
Provider RPM60 / 60
Provider TPM40k / 40k
Done
0
Errors
0
Queued
0
Goodput
—

Two buckets, reserve & reconcile

Admits only when RPM and TPM both have room. Reserves an estimate, refunds the difference.
Provider RPM60 / 60
Provider TPM40k / 40k
Done
0
Errors
0
Queued
0
Goodput
—
Press Run. Both clients get the identical request stream and both queue what they cannot send yet. The only difference is the admission rule.