Deploy and Query a Model for Realtime Inference MACHINE-LEARNING-ASSOCIATE Practice Questions (Page 4)

Questions 16–18

1. 16application · medium

   A model serving endpoint for a recommendation engine is experiencing increased latency during peak hours. The team has already scaled up the endpoint to its maximum number of instances. They suspect the model's feature computation is the bottleneck. What is the most effective next step to improve performance?

   • AIncrease the endpoint's maximum number of instances further.
   • BOptimize the model's feature engineering code to reduce computation time per request.
   • CSwitch the endpoint to provisioned throughput mode.
   • DCache the model's predictions in a Redis-like store.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   When configuring a Databricks Model Serving endpoint, which of the following must be specified?

   • AThe model name and version to serve
   • BThe database name where the model metadata is stored
   • CThe path to a training script in a Git repository
   • DThe cluster ID of the compute that will host the model
   Reveal Solution Discussion Select an answer first
3. 18foundation · easy

   A machine learning engineer needs to update a deployed model endpoint to serve a new version of the model. What is the correct way to do this in Databricks Model Serving?

   • ADelete the existing endpoint and create a new one with the same name
   • BRedeploy the model to a new cluster and update the DNS record
   • CCreate a new endpoint with a different name and update the application to use it
   • DEdit the endpoint configuration to point to the new model version
   Reveal Solution Discussion Select an answer first