DP-900 question patterns

JSON documents (document store such as Cosmos DB).

Schema is carried in the data, not enforced by a table definition.

Azure Blob Storage.

Unstructured data suits object storage.

Transactional relational database such as Azure SQL Database.

Atomicity and isolation protect each transaction.

Data warehouse or lakehouse with a star schema.

Analytical workloads scan many rows; don't run them on the OLTP database.

Parquet (or ORC); Delta adds transactions on top.

Reading a few columns is cheap and values compress well.

Avro.

Row-based binary format common with Event Hubs and Kafka.

Transform first = ETL; load raw then transform in the warehouse or lake = ELT.

ELT uses the scalable target engine and is the cloud norm.

DBA.

Operational care of databases is DBA work.

Data engineer.

Moving and shaping data is engineering.

Data analyst, usually with Power BI.

Analysts model, visualise and explain.

Stored procedure.

Views return data, indexes speed lookups, procedures hold logic.

View.

A view stores a query, not data.

Create an index.

Slows writes slightly and uses storage.

Normalise to 3NF for OLTP.

Each fact stored once; cost is more joins.

Foreign key to the customer primary key.

Enforces referential integrity.

HAVING, not WHERE.

WHERE filters rows before grouping.

Azure SQL Database (PaaS).

Microsoft manages patching, backup and HA.

Azure SQL Managed Instance.

Near 100% SQL Server compatibility, still PaaS.

SQL Server on Azure Virtual Machines (IaaS).

Only option with full OS control.

Azure SQL Database elastic pool.

Databases share one pool of resources.

Serverless Azure SQL Database.

Auto-pauses when idle and bills per second.

Azure Database for PostgreSQL or MySQL flexible server.

Keeps engine and drivers, Microsoft runs the service.

Append blob.

Optimised for append operations.

Archive access tier.

Cheapest storage but offline; needs rehydration.

Blob lifecycle management policy.

Rules on age or last access tier or delete.

Blob storage with hierarchical namespace.

Adds directories, atomic rename and POSIX ACLs.

Azure Files (with File Sync for local caching).

Shares behave like a mapped drive.

Azure Table storage.

Schemaless NoSQL without joins.

Azure Cosmos DB for NoSQL.

Native API with automatic indexing and multi-region writes.

Cosmos DB for MongoDB.

Wire-protocol compatibility.

Cosmos DB for Apache Gremlin.

Graph model with vertices and edges.

Cosmos DB for Apache Cassandra.

Wide-column model using CQL.

Azure Cosmos DB for Table.

Same API, premium capabilities.

Session consistency.

Reads your own writes at good latency.

Read-access geo-redundant storage.

Secondary region is readable.

Zone-redundant storage.

Copies span availability zones.

Microsoft Fabric.

One capacity, one lake, all workloads.

Azure Databricks.

Managed Spark with Delta Lake.

Azure Data Factory or Fabric Data Factory pipelines.

Orchestration and movement, not heavy compute.

Synapse serverless SQL pool or Fabric SQL analytics endpoint.

Pay per query, no cluster.

Batch processing.

Latency of hours is acceptable.

Stream processing with Event Hubs and Stream Analytics.

Value decays quickly.

Azure Event Hubs.

Big-data ingestion broker, Kafka-compatible.

Azure Stream Analytics.

Managed real-time query engine.

Fabric Eventhouse or Azure Data Explorer.

KQL-optimised store for telemetry.

Power BI Desktop, then publish to the service.

Service is for sharing, dashboards and refresh.

Power BI dashboard.

Reports are multi-page from one model.

DAX measure.

Calculated columns are stored per row at refresh.

Line or area chart.

Best for continuous time series.

Bar chart for categories; scatter plot for relationship.

Match visual to the question.