Databricks-ML-Pro Model Deployment Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Set the endpoint's scale-to-zero behavior so that instances remain warm and are not fully shut down during idle periods.

  Why this is correct

  Scale-to-zero controls whether the endpoint's compute is torn down when there is no traffic. Disabling it, or configuring the endpoint to keep instances provisioned, means the model stays loaded and the first request after idle time does not pay the container start and model load cost, which is the dominant contributor to cold-start latency.

• ✓

  Reduce the size of the model artifact and its logged dependencies so that container startup and model loading complete faster.

  Why this is correct

  Cold-start time is largely the sum of provisioning a container, installing the environment, and loading the model artifact into memory. A smaller artifact and leaner dependency set shrink the model-loading portion directly, so the endpoint reaches a ready state sooner and the first request is served with less delay.

• ✗

  Enable Inference Tables on the endpoint so that request payloads are cached and replayed on subsequent cold starts.

  Why it's wrong here

  Inference Tables log request and response payloads to a Delta table for audit and monitoring. They are a logging sink, not a cache, and nothing about them preloads a model or replays requests to warm an endpoint, so they have no effect on cold-start latency.

• ✗

  Increase the number of served model versions on the endpoint so that requests can be spread across more copies of the model.

  Why it's wrong here

  Adding served model versions creates additional distinct models or versions behind the endpoint; it does not add parallel replicas of the same model for load spreading. Each version still needs its own warm start, and the endpoint routes by traffic configuration rather than duplicating a single version for latency reduction.

• ✗

  Lower the endpoint's concurrency setting so that each replica handles fewer simultaneous requests and starts faster.

  Why it's wrong here

  Concurrency determines how many requests a single replica processes in parallel; lowering it reduces throughput per replica and can actually increase queueing and perceived latency. It does not shorten the time required to provision the container and load the model, so it does not address cold start.