NVIDIABased on real interviews
Investigate a slow experiment dashboard
Locate the delay between a finished experiment and the metrics appearing in its dashboard.
Unlock with Full AccessThe challenge
The training experiment finishes, but its metrics take several more minutes to appear in the dashboard. Results pass through a client, API, Kafka consumers, storage and a polling browser. Logging volume recently increased, and a teammate wants to add Kafka capacity before those stages have been timed.
Use the supplied path to plan an investigation of one metric batch from send to display. Each proposed check should narrow the explanation or change the next action, so the team can distinguish a capacity problem from delays elsewhere in the path.
What you'll practise
Latency decomposition
Determine where time is spent across the delivery and aggregation path.
Evidence before scaling
Distinguish a capacity issue from other causes of delayed results.
Explain which evidence you would inspect and what each result would change.