Load control

Changes apply immediately to every request in flight.

Live

Median latency

—ms

p95 —

Queue depth

—

limit —

Busy workers

—

of — slots

Dropped / failed

—

— completed

Server capacity

Worker slots 4

How many requests the server can process at the same time. Fewer slots means longer queues.

Work per request 150 ms

How long a single request occupies a slot. Higher means every visitor waits longer, even when capacity is fine.

Queue limit 40

Once this many requests are waiting, new ones are rejected immediately instead of queued.

Background visitors 8 /s

Synthetic visitors generated by the server itself. This is what creates genuine overload without you clicking anything.

CPU burn per request 4 ms

Extra synchronous CPU work blocking the event loop. Adds real server strain, not just delay.

Error rate 0 %

Percentage of requests that fail with an upstream error once they reach a worker.

Load simulation active

Turn this off to serve instantly, useful for comparing a fast site against a slow one.

One click applies a full profile

All changes saved

Latency over the last minute —