You work for an advertising company, and you've developed a Spark ML model to predict click-through rates at advertisement blocks. You've been developing everything at your on-premises data center…

Question

Options

• AUse Cloud ML Engine for training existing Spark ML models
• BRewrite your models on TensorFlow, and start using Cloud ML Engine
• CUse Cloud Dataproc for training existing Spark ML models, but start reading data directly from BigQuery
• DSpin up a Spark cluster on Compute Engine, and train Spark ML models on the data exported from BigQuery

• A
  13% (8)
• B
  6% (4)
• C
  77% (48)
• D
  3% (2)

Explanation

Option C is correct because Cloud Dataproc is Google Cloud's managed Spark service, making it the perfect lift-and-shift destination for existing Spark ML code - no rewrites needed. Dataproc also has a native BigQuery connector, so training pipelines can read data directly from BigQuery without extra export steps.

Why the distractors are wrong:

• A - Cloud ML Engine (now Vertex AI) does not support Spark ML natively; it's built for TensorFlow, scikit-learn, and XGBoost.
• B - Rewriting models in TensorFlow directly contradicts the "rapid lift-and-shift" requirement; this would be a significant re-engineering effort.
• D - Manually managing a Spark cluster on Compute Engine works but is inefficient and unmanaged; Dataproc does the same thing with far less operational overhead. Exporting data out of BigQuery also adds unnecessary steps when direct reads are available.

Memory tip: On the exam, whenever you see Spark + GCP, think Dataproc - it's the managed Spark service that enables true lift-and-shift. "Lift-and-shift = keep the framework, move the infrastructure."

No community discussion yet for this question.