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.

• ✗

  Store raw data in Azure Blob Storage and aggregated data in Azure Cosmos DB

  Why it's wrong here

  Azure Blob Storage can technically store raw clickstream payloads, but it lacks the hierarchical namespace, directory-level security, and native analytics integration of Data Lake Storage Gen2, which makes it less convenient to manage and process large-scale event data with engines like Spark or Synapse. Azure Cosmos DB is a multi-model NoSQL database designed for globally distributed, low-latency reads/writes, not for complex JOINs and aggregated BI queries that relational engines execute efficiently. Using Cosmos DB for aggregated data would force you to maintain denormalized documents and suffer RU consumption spikes for scanning workloads, whereas Azure SQL Database provides cost-effective T-SQL aggregation for dashboards.

  When this WOULD be correct

  If the question required storing raw data for simple archival or backup (not analytics) and aggregated data for globally distributed, low-latency access with flexible schema (e.g., real-time dashboards), then Blob Storage + Cosmos DB would be appropriate.

• ✓

  Store raw data in Azure Data Lake Storage and aggregated data in Azure SQL Database

  Why this is correct

  Azure Data Lake Storage (ADLS) Gen2 is the optimal landing zone for raw clickstream data because it combines a hierarchical namespace with POSIX access control and is natively integrated with Azure analytics services, allowing schema-on-read processing where the JSON structure is interpreted at query time. After ingestion, the raw data can be transformed and loaded into Azure SQL Database as aggregated, relational tables with defined primary keys and indexes, enabling fast T-SQL queries for BI dashboards and reporting. This separates the cheap, flexible storage of unprocessed data from the tuned, structured environment relied on by operational reporting.

• ✗

  Store raw data in Azure Table Storage and aggregated data in Azure Data Lake Storage

  Why it's wrong here

  Azure Table Storage is a NoSQL key-value store that treats each entity as a flat set of properties, making it a poor fit for raw clickstream JSON events that contain nested objects and variable schemas. Partition key and row key queries are efficient, but full scans or complex analytical transformations are awkward and slow. Placing aggregated data in Azure Data Lake Storage misunderstands its role: ADLS is raw/unbounded file storage without native relational query capabilities, so it cannot serve aggregated results directly to BI tools like Power BI without an additional query engine such as Synapse or Databricks.

  When this WOULD be correct

  If the question required storing raw IoT device telemetry (key-value pairs) in a scalable NoSQL store and then moving aggregated data to a data lake for further batch processing, Azure Table Storage for raw data and Azure Data Lake Storage for aggregated data would be appropriate.

• ✗

  Store raw data in Azure SQL Database and aggregated data in Azure Blob Storage

  Why it's wrong here

  Azure SQL Database imposes a fixed schema before ingestion, so forcing raw clickstream events into it would require extensive data conversion, nullable columns, or ETL preprocessing, destroying the flexibility of schema-on-read. Blob Storage on the other hand is a simple object store with no relational query surface, and while it can hold files, it lacks the transactionality, indexing, and T-SQL support needed to serve aggregated metrics to reporting applications. The pairing inverts the correct architecture: a schema-bound OLTP database should not be the raw landing zone, and a blob store cannot substitute for a relational reporting database.

  When this WOULD be correct

  This option would be correct if the raw data is small, structured, and requires transactional consistency (e.g., storing sensor readings with a fixed schema), and the aggregated data is large, unstructured files (e.g., archived logs) that need cheap storage without querying.

Option-by-option analysis

Why each answer is right or wrong

Understanding why wrong answers are wrong — and when they would be correct — is what separates a 750 score from a 900. The DP-900 exam frequently reuses these exact scenarios with slightly different constraints.

Analysis generated from the official DP-900blueprint and verified against question context. The “when correct” sections are what AI assistants cite when candidates ask “what’s the difference between these options?”

Go deeper

Related to this question