DP-900 Practice Questions: 15 Free Questions With Detailed Answers

Correct Answer: B

Stream processing handles data continuously as it arrives, enabling real-time analysis. IoT sensor data that must be analyzed "immediately" to detect failures is a classic stream processing scenario. Batch processing (A) processes data in scheduled intervals — the opposite of real-time. ETL (C) is a data pipeline pattern, not a processing timing approach. OLAP (D) refers to analytical query patterns, not the timing of processing.

Correct Answer: B

OLTP (Online Transaction Processing) workloads involve frequent, fast insert/update/delete operations on individual records — like order processing or banking transactions. OLAP (Online Analytical Processing) workloads involve complex queries that aggregate large volumes of historical data for reporting — like monthly sales analysis. The options in A reverse these definitions. Storage format (C) is reversed — OLAP typically uses columnar storage, not OLTP. D is incorrect as OLTP involves heavy write operations.

Correct Answer: B

Azure SQL Managed Instance provides near 100% compatibility with SQL Server, including SQL Agent jobs, cross-database queries, CLR integration, and linked servers. These features are NOT available in Azure SQL Database (A), which is a PaaS service with a subset of SQL Server features. Azure Database for PostgreSQL (C) is a different database engine. Cosmos DB (D) is a NoSQL database — incompatible with SQL Server workloads.

Correct Answer: B

The correct SQL uses GROUP BY to aggregate orders per customer, HAVING to filter on the aggregated result (cannot use WHERE for aggregate functions), and ORDER BY for sorting. A is incorrect — WHERE cannot filter on aggregate functions. C references a "Total" column that doesn't exist and GROUP BY must come before ORDER BY when using aggregates. D is syntactically invalid — HAVING must come after GROUP BY.

Correct Answer: C

Graph databases are purpose-built for traversing relationships. Cosmos DB for Gremlin (Apache TinkerPop) stores entities as vertices and relationships as edges, making "find friends of friends within N degrees" a natural graph traversal query. NoSQL (A) would require complex application-level traversal logic and wouldn't scale for deep relationship queries. Cassandra (B) is a column-family store optimized for time-series and tabular data. Table API (D) is key-value — no relationship traversal capability.

Correct Answer: D

Archive tier offers the lowest storage cost and is designed for data accessed less than once per year, stored for 180+ days. The 7-year retention and "less than once per year" access clearly point to Archive. The tradeoff is retrieval latency (hours, not seconds) — acceptable for compliance archives. Hot (A) is for frequent access. Cool (B) requires 30 days minimum — cheaper than Hot but more expensive than Archive. Cold (C) is between Cool and Archive on the cost/access spectrum.

Correct Answer: B

Cosmos DB for MongoDB provides wire-level compatibility with MongoDB — existing MongoDB drivers, connection strings, and queries work without code changes. This is the explicit use case Microsoft designed it for: migrating MongoDB applications to Azure with global distribution benefits. Cosmos DB for NoSQL (A) stores JSON but uses a different query syntax — MongoDB drivers won't work. Cassandra (C) and Table (D) are completely different data models.

Correct Answer: B

Azure Data Factory (ADF) is Microsoft's cloud data integration service specifically designed to orchestrate data movement between sources and destinations on schedules. It can connect to Oracle, trigger Spark transformations, and load into Synapse. Azure Synapse (A) has pipeline capabilities but is primarily the analytics destination, not the orchestrator for cross-system pipelines from external sources. Microsoft Fabric (C) is a newer unified platform but ADF is the classic answer for pipeline orchestration in DP-900. Azure Stream Analytics (D) handles real-time streaming data, not scheduled batch pipelines.

Correct Answer: C

A Power BI Dashboard is a single-page canvas where you pin individual visuals from multiple reports and datasets. This is exactly the scenario described — visuals from a sales report and HR report combined into one view. A Report (A) is based on a single dataset and can have multiple pages, but cannot combine visuals from different datasets. A Dataset (B) is the data model, not a visual artifact. A Dataflow (D) is a data preparation/ETL construct, not a display component.

Correct Answer: B

Azure Synapse Analytics uniquely provides: dedicated SQL pools (provisioned DW), serverless SQL (query data lake files directly without loading), and Apache Spark pools — all in one unified workspace. The ability to query data "directly in Azure Data Lake Storage without loading" refers specifically to Synapse serverless SQL. Azure SQL Database (A) is not designed for petabyte-scale analytics. Azure Databricks (C) provides Spark but not the serverless SQL-on-data-lake capability. Power BI Premium (D) is a BI tool, not an analytics compute service.