Design data integration

Rookwood Logistics has a team of data scientists who build machine learning models in Python and Scala notebooks over 200 TB of Parquet files in Azure Data Lake Storage Gen2. They need collaborative notebooks, Delta Lake tables, MLflow experiment tracking, and the ability to run the latest open-source Apache Spark runtime with GPU-enabled clusters that scale automatically. The business analysts have a separate solution and are out of scope.

Which analytics platform should you recommend?

1. AAzure Synapse Analytics dedicated SQL poolA dedicated SQL pool is a massively parallel T-SQL data warehouse; it does not run Spark notebooks, Delta Lake, or MLflow.
2. BAzure DatabricksDatabricks is the Spark-native platform in Azure with collaborative notebooks in Python, Scala, SQL, and R, native Delta Lake, managed MLflow, GPU-enabled autoscaling clusters, and the most current Spark runtimes.
3. CAzure Data Explorer clusterData Explorer is a columnar engine for fast interactive queries over telemetry using KQL; it is not a Spark or machine learning training platform.
4. DAzure Synapse Analytics serverless SQL poolServerless SQL pools query files in the lake with T-SQL and bill per query, which suits ad hoc exploration but provides no notebooks, Delta writes, or ML experiment tracking.

Skyvault Networks ingests 5 TB per day of firewall and application logs from Azure Event Hubs. Security analysts need to run ad hoc, interactive queries over the last 90 days that return in seconds, including free-text search, time-series aggregation, and anomaly detection, and they want to keep two years of data queryable at lower cost. The data is append-only and schema changes frequently. Skyvault wants a managed service that minimizes administrative effort.

Which analytics solution should you recommend?

1. AAzure Data Explorer with hot cache set to 90 days and a two-year retention policyData Explorer is purpose-built for high-volume append-only telemetry: it ingests natively from Event Hubs, indexes free text, provides time-series and anomaly detection functions in KQL, and separates a hot SSD cache period from a longer, cheaper retention period.
2. BAzure Synapse Analytics dedicated SQL pool loaded by Azure Stream AnalyticsA dedicated SQL pool is a relational warehouse that requires a defined schema and tuned distributions; free-text search and anomaly detection are not native, and keeping two years of 5 TB per day resident is expensive.
3. CAzure Databricks streaming jobs writing Delta tables, queried with Spark SQLDatabricks can handle the volume, but interactive sub-second queries require clusters kept running and tuned, and the analysts would need to build text search and anomaly detection themselves, which increases administrative effort.
4. DAzure Cosmos DB for NoSQL with the change feed and analytical storeCosmos DB is an operational database; ingesting 5 TB per day into it is costly in request units, and it is not designed for ad hoc analytical scans with free-text search.

Welbrook Retail needs to copy nightly sales extracts from an on-premises Oracle database and from a SaaS CRM REST API into Azure Data Lake Storage Gen2, apply joins and aggregations without writing code, and orchestrate the steps with retries and alerts. The on-premises network allows only outbound connections. Welbrook wants a managed service and minimal custom code.

Which two actions should you perform? Each correct answer presents part of the solution.

1. ACreate an Azure Data Factory with pipelines that use mapping data flows for the joins and aggregationsData Factory is the managed orchestration and ETL service with built-in connectors for Oracle, REST, and Data Lake Storage Gen2, and mapping data flows provide visually designed transformations executed on managed Spark without code.
2. BInstall a self-hosted integration runtime on a server in the on-premises network to reach the Oracle databaseA self-hosted integration runtime runs inside the private network and makes only outbound connections to Data Factory, which lets pipelines read on-premises data sources without inbound firewall rules.
3. CUse Azure Data Share to publish the Oracle data to Welbrook's Data Lake Storage accountData Share shares data between Azure tenants from supported Azure sources; it cannot pull from an on-premises Oracle database or a REST API and performs no transformations.
4. DUse an Azure-hosted integration runtime with a managed virtual network to reach the Oracle databaseAn Azure integration runtime, even with a managed virtual network and private endpoints, cannot connect to an on-premises server that accepts no inbound connections.
5. EWrite Azure Functions in Python that query Oracle and the REST API and write Parquet filesFunctions could do the job but require custom code for connectivity, transformation, retries, and alerting, which contradicts the minimal custom code requirement.

Amberfield Insurance curates a set of claims statistics in Azure Data Lake Storage Gen2 and in Azure Synapse Analytics dedicated SQL pool tables. Twelve partner reinsurers in separate Microsoft Entra tenants need regularly refreshed copies of these datasets in their own subscriptions. Amberfield must be able to see which partner has accepted the data, update the shared data on a schedule, and revoke a partner's access at any time, with no pipelines built or maintained by the partners.

Which solution should you recommend?

1. AAzure Data Share with snapshot-based sharing to each partnerData Share is designed for cross-tenant data sharing: the provider creates a share from supported sources such as Data Lake Storage Gen2 and Synapse, invites recipients, schedules incremental snapshots, tracks acceptance, and can revoke a share at any time.
2. BAn Azure Data Factory pipeline per partner that copies data into each partner's storage accountThis requires Amberfield to hold credentials or managed identity access into every partner tenant and to maintain twelve pipelines, with no built-in invitation, acceptance tracking, or revocation model.
3. CShared access signature URLs to the Data Lake Storage containers emailed to each partnerSAS tokens give direct read access but do not cover the Synapse tables, provide no acceptance tracking, and revocation requires rotating keys or stored access policies for all partners at once.
4. DAzure Synapse Link enabled for each partner's workspaceSynapse Link replicates operational data from sources such as Cosmos DB or SQL into Synapse within one organization; it is not a cross-tenant sharing mechanism.