Attached Synapse Spark pool is the correct choice for the first activity because it integrates Azure ML with Azure Synapse Analytics, a platform optimized for processing "vast amounts of data," managing "linked services," and natively integrating with Microsoft Power BI. Serverless Spark compute is correct for the second activity because it provides on-demand compute without requiring a dedicated cluster setup. It is specifically recommended for "ad hoc" interactive work and uses the "fewest resources" by scaling to zero when not in use and requiring no infrastructure management overhead.

Microsoft Learn. "Apache Spark in Azure Machine Learning". Concepts.

Reference: "Serverless Spark compute works well for most user scenarios that require quick access to distributed computing resources... No dependencies on creation of other Azure resources." (Section: Points to consider).

Reference: "Attached Synapse Spark pool... If you already have an Azure Synapse Analytics workspace... you can attach it to your Azure Machine Learning workspace." (This supports the scenario involving established pipelines and linked services).

URL: https://learn.microsoft.com/en-us/azure/machine-learning/apache-spark-azure-ml-concepts

Microsoft Learn. "Interactive data wrangling with Apache Spark in Azure Machine Learning". How-to Guide.

Reference: "Use of a serverless Spark compute is the easiest way to access a Spark cluster in Azure Machine Learning... The serverless Spark compute doesn't require creation of resources in the Azure Synapse workspace."

URL: https://learn.microsoft.com/en-us/azure/machine-learning/interactive-data-wrangling-with-apache-spark-azure-ml

Microsoft Learn. "What is Azure Synapse Analytics?". Overview.

Reference: Confirms Synapse's role in "ingesting" vast data and its built-in integration with "Power BI" and "Linked Services".

URL: https://learn.microsoft.com/en-us/azure/synapse-analytics/overview-what-is