Task 2.3: Design data integration

• AAn Azure Data Factory pipeline that uses only the Azure integration runtime

  Incorrect

  The Azure integration runtime runs in Azure and can't reach a server behind the corporate firewall without a private network path.

• BAn Azure-SSIS integration runtime that runs new SSIS packages

  Incorrect

  Building new packages is the custom development the requirement avoids, and the copy still needs a path to the private server.

• CAn Azure Data Box shipped every night

  Incorrect

  Shipping a device every night can't deliver a scheduled, centrally monitored copy each morning.

• DAn Azure Data Factory pipeline that uses a self-hosted integration runtime on-premises

  Correct

  A self-hosted integration runtime reaches private networks with outbound connections only, and the pipeline schedules and monitors the copy.

The self-hosted integration runtime is installed inside the private network and opens only outbound connections to Data Factory. Copy activities then move data from on-premises stores to Azure on a schedule, with Data Factory monitoring every run.

• AAn Azure Data Factory pipeline triggered every five minutes

  Incorrect

  Data Factory moves data in batches on triggers; it isn't a continuous stream processor with sub-minute alerts.

• BAn Azure Stream Analytics job with windowing functions

  Correct

  Stream Analytics runs SQL-like queries over event streams, with tumbling, hopping and sliding windows, and writes to SQL Database.

• CA Logic Apps workflow that runs for each event

  Incorrect

  A workflow per event at this volume is costly, and windowed aggregates would need custom state handling.

• DAn Azure Batch pool that processes the event backlog each hour

  Incorrect

  Hourly batches miss the per-minute averages and second-level alerts.

Azure Stream Analytics is a real-time analytics service that reads from Event Hubs and other inputs and uses a SQL-like query language with temporal windowing functions to aggregate events and write results to outputs such as Azure SQL Database.

• AA Data Factory copy pipeline scheduled every hour into a lakehouse

  Incorrect

  An hourly pipeline is the ETL work the team wants to avoid and delivers data up to an hour old.

• BA readable geo-secondary of the database for the analysts

  Incorrect

  A geo-secondary keeps reporting off the primary but isn't data in Fabric's OneLake for Fabric analytics.

• CMirror the Azure SQL database into Fabric OneLake

  Correct

  Mirroring replicates the database into OneLake in near real time as Delta tables, with no pipelines to build, and analysts query the copy.

• DA nightly BACPAC export of the database to Data Lake Storage

  Incorrect

  A nightly export is a day old, and BACPAC files aren't queryable tables.

Mirroring in Fabric continuously replicates an operational database, such as Azure SQL Database, into OneLake as Delta tables in near real time, so analytics run on the replica without ETL pipelines.

• AA dedicated SQL pool loaded with the Parquet data

  Incorrect

  A dedicated pool reserves compute that is billed while it runs, and the data would have to be loaded first.

• BThe workspace's built-in serverless SQL pool

  Correct

  Serverless SQL pool queries lake files with T-SQL, needs no infrastructure, and bills by the data processed.

• CAn Azure SQL Database in Business Critical with the files loaded into tables

  Incorrect

  A provisioned database bills compute all the time for a few weekly queries.

• DAn Azure HDInsight cluster running Hive

  Incorrect

  An HDInsight cluster is billed while it exists and isn't T-SQL over the lake.

Every Synapse workspace has a serverless SQL pool endpoint. It queries Parquet, CSV and Delta files in the lake with T-SQL, there's nothing to provision, and you pay for the data processed by each query.

• AAzure HDInsight with Spark clusters

  Incorrect

  HDInsight runs open-source Spark but not Databricks notebooks, MLflow tracking or Unity Catalog as they are.

• BAzure Databricks

  Correct

  Azure Databricks is the Databricks platform on Azure, with notebooks, MLflow and Unity Catalog.

• CAzure Data Factory mapping data flows

  Incorrect

  Mapping data flows are visual transformations, not a notebook and ML platform.

• DAzure Stream Analytics jobs

  Incorrect

  Stream Analytics jobs don't run notebooks or track machine learning experiments.

Azure Databricks is a unified analytics platform for data engineering, data science and machine learning, including notebooks, MLflow and Unity Catalog for governance.

• AAzure SQL Database Hyperscale

  Incorrect

  Hyperscale is an OLTP database queried with T-SQL, not a KQL analytics engine for billions of records a day.

• BAzure Cosmos DB for NoSQL

  Incorrect

  Cosmos DB serves operational point reads and writes, not interactive analytics over months of telemetry.

• CAzure Queue Storage with a nightly batch

  Incorrect

  Queues pass messages between components; they don't analyze data.

• DAzure Data Explorer

  Correct

  Azure Data Explorer is a fast analytics service for large volumes of telemetry and time series, queried with KQL.

Azure Data Explorer is a fully managed, high-performance analytics platform for large volumes of data in near real time, with KQL and built-in time-series analysis and machine learning functions.

• AA function triggered by Event Hubs that writes each event to a blob

  Incorrect

  A function is a consumer app the team would have to write and run.

• BEvent Hubs Capture to Data Lake Storage

  Correct

  Capture writes streaming data to Blob Storage or Data Lake Storage on a time or size window, with no code to run.

• CEvent Grid subscriptions that push each event to storage

  Incorrect

  Event Grid doesn't read an event hub's stream or batch it into files in the lake.

• DA larger retention period on the event hub

  Incorrect

  Retention keeps events in the hub for a limited time; it doesn't land files in the data lake.

Event Hubs Capture automatically delivers streaming data to Azure Blob Storage or Azure Data Lake Storage in Avro or Parquet files, on a time or size interval that you choose.

• AAzure Resource Graph queries over the data resources

  Incorrect

  Resource Graph doesn't catalog datasets, their classification or lineage, or on-premises sources.

• BTags on each storage account and database

  Incorrect

  Tags describe resources, not the datasets inside them, and don't show lineage.

• CMicrosoft Purview Data Map and Unified Catalog

  Correct

  The Data Map scans hybrid and multicloud sources with built-in classification and lineage, and Unified Catalog lets people discover the data.

• DMicrosoft Defender for Cloud regulatory compliance

  Incorrect

  Regulatory compliance assesses resource configuration against standards, not data lineage.

The Microsoft Purview Data Map scans and classifies data across hybrid, on-premises and multicloud systems and captures lineage, and Unified Catalog builds on it so users can find and understand data.