—
INTERACTIVE EXPERIMENT / A TEACHING MODEL
Beef Noodle Benchmark
How to play
Start with one bowl per batch, then change just the batch size. Compare the first customer’s wait with the whole group’s completion time.
- Choose a scenario and press “Open shop”. Pause, resume or skip to the result at any time.
- Select a customer to see batching wait, pot wait and cooking time. Their total is arrival-to-bowl latency.
- Finish a run, change one setting and try again. Compare completed runs with the same arrival pattern.
Batch = bowls cooked together. Latency = one customer’s time from arrival to bowl. Throughput = bowls served per minute since opening.
—
—
Not here yet
The clock starts when the shop opens.
No customers here.
No customers here.
No customers here.
No customers here.
Model, measurements and limits
A deliberately simplified, reproducible scheduling model. Its numbers explain batching and measurement; they do not predict restaurant capacity or CPU/accelerator performance.
- Eight customers, one bowl each, first-in-first-out. A single pot handles one batch at a time. Everyone in a batch receives their bowl together.
- Each batch takes 30 + 6 × actual bowls (seconds), including batch setup. Ordering, payment, mistakes and parallel staff are outside the model.
- A full batch starts immediately when the pot is free. Otherwise the oldest waiting customer’s arrival plus the batching limit sets the earliest start. A busy pot can delay it further.
- At simultaneous events, serve completed bowls, admit new arrivals, then dispatch the next batch. Animation frame rate does not decide the schedule.
- Latency is each customer’s completion minus their own arrival time. The mean includes served customers only; waiting customers keep their visible timers.
- Throughput is bowls served divided by time since opening, shown as bowls/minute. Time starts at zero and includes setup and idle periods. This is an observed run average, not steady-state capacity.
- Batching wait is time queued while the pot is idle; pot wait is time queued while it is busy. Add both to cooking time to get that customer’s latency.
- Amortized time appears only after all bowls are ready: total elapsed time divided by total bowls. It is not a customer’s latency.
- Playback speed and skipping change viewing progress, not the model. Different arrival patterns are not directly comparable rankings.