4. Non-Relational Data on Azure (15-20%)

Not all data fits into tables. This phase covers Azure Storage — Blob, Files, Table and Queue — and Azure Cosmos DB.

Why this matters: this is the smallest domain by weight, but the one where a wrong choice is most expensive in practice. Putting video files in a database, or mounting a file share to serve a website, are mistakes that survive into production and cost real money before anyone notices.

💡 The mental model: one storage account offers four services, each matching an access pattern rather than a data type. Blob — objects fetched by URL. Files — a share you mount at a drive letter or path. Table — key-value lookup by partition and row key. Queue — messages passed between components. Cosmos DB sits above all four for globally distributed operational data with latency guarantees.

⚠️ The trap: the discriminator is almost never what the data is — it is how the application reaches it. The same JPEG belongs in Blob Storage if a browser requests it over HTTP, and in Azure Files if a legacy application opens it through a UNC path. When a question describes a file, do not stop at "unstructured"; find the sentence that says how something gets to it.