DP-900 Describe core data concepts 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.

• ✗

  Keep current pipeline but replace Synapse with Azure Analysis Services for faster query performance.

  Why it's wrong here

  Replacing the serving engine with Azure Analysis Services would only accelerate interactive queries over already-loaded data; it does nothing to shorten the ingestion and transformation steps upstream. AAS requires a refreshed tabular model—populated from the Data Lake or Synapse—so dashboard freshness remains bounded by the existing batch cadence. Because the 10-minute requirement is an end-to-end freshness constraint, a faster query engine alone cannot close the gap.

• ✗

  Use Azure Data Factory with tumbling window triggers every 5 minutes to load data from Data Lake to Synapse.

  Why it's wrong here

  A Data Factory tumbling window trigger at 5-minute intervals still represents scheduled batch orchestration, not event-driven streaming. Every window adds at least a five-minute delay from data arrival to pipeline start, and the actual Copy activity into Synapse introduces additional execution time that can easily push end-to-end latency beyond 10 minutes. It also creates overlapping windows and retries, which increase operational complexity but do not deliver the low-latency, continuous processing required.

• ✓

  Ingest data into Azure Event Hubs, use Azure Stream Analytics to process and output to Power BI for real-time dashboards.

  Why this is correct

  Event Hubs captures telemetry continuously and Stream Analytics runs persistent SQL-style queries over that stream, producing low-latency results directly in Power BI's streaming dataset. Because processing begins as events arrive and there is no dependency on file batches, dashboards typically reflect updates within seconds rather than minutes. This satisfies the 10-minute freshness SLA and provides the real-time visualization experience the business expects.

• ✗

  Increase the number of Databricks clusters and use Auto Loader to speed up transformations.

  Why it's wrong here

  Scaling Databricks clusters and using Auto Loader only accelerates the existing file-based batch processing; Auto Loader detects new files in object storage and triggers structured streaming micro-batches, but the pipeline is still constrained by how often files appear and how long each batch takes to write to the serving layer. More workers may shrink processing duration, but the design remains batch-to-batch and cannot guarantee bounded sub-10-minute latency during traffic spikes. The requirement demands a streaming-first architecture, not faster compute on the same data movement path.

Go deeper

Related to this question