Databricks Certified Machine Learning Professional Practice Questions

Databricks Certified Machine Learning Professional exam preparation

Databricks Certified Machine Learning Professional practice questions in English: 180 questions with explanations. The mock exam has 59 questions and 120 minutes.

Databricks Certified Machine Learning Professional free question 3 / 10 · Domain 1: Model Development

A training workflow runs a heavy Spark ETL step and then a Ray training step in one task on one cluster. The two engines compete for the same memory and CPUs. Which workflow pattern from the Databricks guidance reduces this contention?

  1. Run the ETL step on the Ray head node and leave the workers free.
  2. Start the Ray cluster with a larger head node and keep one task.
  3. Isolate the ETL in its own subtask, on a cluster matched to ETL.
  4. Run the ETL step inside a Ray task through a Spark UDF call.
Show the answer and explanation

Answer: C

Why this is correct
The guide recommends isolating and separating the main extract-transform-load portion into its own subtask within a Databricks Workflow. That lets the team match the cluster type to the type of ETL workload and avoids resource sharing issues between Ray and Spark. The Ray step then runs as its own task and reads the prepared data. When Ray and Spark must share a single task, the guide offers other patterns, with Spark for data handling and Ray for computation.

Why the other options are wrong
"Run the ETL step inside a Ray task through a Spark UDF call." … Running Ray inside Spark functions is an advanced pattern in the guide, and running the ETL inside Ray does not separate the resources of the two engines.
"Run the ETL step on the Ray head node and leave the workers free." … By default the Ray head node is limited to 0 CPU cores and 128 MB of memory because it only coordinates; it is not a place for ETL.
"Start the Ray cluster with a larger head node and keep one task." … A larger head node does not separate the engines. The guide recommends isolating the ETL so that the cluster type can match the ETL workload.

Separate the engines when they compete
If both must run in one task, resource limits can be set when the Ray cluster starts.

Go to the Databricks Certified Machine Learning Professional practice app (10 free questions; purchase unlocks all questions)

See the full practice set