Describe an Analytics Workload on Azure
The shape of an analytics solution
Analytics workloads ingest data, store it at scale, transform it, and serve it to reports or models. Typical Azure components:
- Azure Data Lake Storage Gen2 - hierarchical namespace over Blob storage; the landing zone for raw files (often organized in a medallion pattern: bronze raw, silver cleansed, gold curated).
- Azure Data Factory - orchestration and data movement (pipelines, activities, linked services, integration runtimes).
- Azure Synapse Analytics - an integrated workspace with pipelines, a dedicated SQL pool (provisioned MPP data warehouse), a serverless SQL pool (query files in the lake on demand, pay per query), and Spark pools.
- Azure Databricks - managed Apache Spark for engineering and machine learning.
- HDInsight - managed open-source big-data clusters (Spark, Hadoop, Kafka, HBase) when you need specific frameworks.
- Azure Stream Analytics / Event Hubs - real-time ingestion and stream processing.
- Power BI / Azure Analysis Services - semantic models and visualization.
Modeling and traps
Warehouses usually use a star schema: fact tables hold numeric measures and foreign keys; dimension tables hold descriptive attributes. Compare ETL vs ELT (ELT is common in cloud warehouses because storage and compute scale independently) and batch vs streaming (streaming for continuous events, batch for scheduled loads).
Traps: calling a data lake a data warehouse (a lake stores raw files; a warehouse stores modeled, query-ready tables); assuming serverless SQL pools hold data (they query external files); forgetting that data movement in Data Factory requires an integration runtime.
Worked example: the business needs dashboards over years of sales history from many sources, refreshed nightly. Pattern: Data Factory loads raw files into ADLS Gen2, Spark transforms them into star-schema tables in a dedicated SQL pool, and Power BI reports on top.