Latency Troubleshooting Quiz

5 questions Pass: 70% +25 pts (50 with Pro)

Quiz covering Troubleshooting ML

What you'll be asked

  1. When troubleshooting a sudden spike in model inference latency, which of the following is the most immediate first step to isolate the issue?
  2. You observe that inference latency is high only for a specific subset of requests. Which tool is best suited for identifying if the bottleneck is in the pre-processing pipeline or the model forward pass?
  3. What is a common symptom that indicates your model inference latency is being caused by 'data bloat' or excessively large input payloads?
  4. If your model inference latency follows a 'sawtooth' pattern that correlates with garbage collection cycles, what is the most likely cause?
  5. You have a model deployed on a GPU. The model throughput is low, but the GPU utilization is consistently at 100%. What is the most likely bottleneck, and how should you address it?

Answers and explanations come after you submit.