DP-900 Describe core data concepts Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Storing IoT telemetry data with low latency

  Why this is correct

  Cosmos DB is engineered for single-digit-millisecond read and write latency at any scale, making it ideal for high-velocity IoT telemetry where sensors and devices emit continuous time-series data. Its automatic horizontal partitioning distributes partitions across physical machines to absorb massive ingest rates, while multiple consistency levels allow you to tune performance versus data freshness. Global distribution lets telemetry be replicated to edge regions for local access, so device data is both durable and quickly retrievable without operational overhead.

• ✗

  Relational OLTP with complex joins

  Why it's wrong here

  Cosmos DB is a NoSQL database that uses a schema-agnostic data model (documents, graphs, key-value, wide-column) and does not perform relational-style joins across collections efficiently. An OLTP system with complex join-heavy transactions requires enforcing referential integrity and ACID guarantees across normalized tables, which is the purpose of Azure SQL Database rather than Cosmos DB. While the SQL API lets you query JSON documents with some joins within a single document, multi-document set operations must be manually implemented via denormalized embedded data or client-side logic, so it is not a valid substitute for a relational OLTP engine.

• ✓

  Session state management for web applications

  Why this is correct

  Session state storage is a common Cosmos DB workload because each HTTP request can read and write a small session document with consistent, single-digit-millisecond latency. Cosmos DB supports TTL to automatically expire stale session records, preventing unbounded growth, and its global replication enables session access from any geographic region without sticky load-balancer affinity. The official Azure Cosmos DB session-state provider for ASP.NET Core serializes user session data into a JSON document, taking advantage of point reads by session ID instead of expensive scans.

• ✓

  Personalization and recommendation engines

  Why this is correct

  Personalization and recommendation engines thrive on Cosmos DB because it can ingest near-real-time user clickstreams, behavioral events, and profile updates while simultaneously serving read-heavy recommendation queries with global low latency. The store allows denormalized per-user preference and interaction documents to be partitioned by customer ID, giving rapid point reads even as user bases scale into the billions. Cosmos DB feeds into Azure Machine Learning or Azure Synapse pipelines for scoring, returning recommendations without requiring a separate caching tier for hot user data.

• ✗

  Storing large files like images and videos

  Why it's wrong here

  Cosmos DB document items have a maximum size of 2 MB, so large binary assets such as images and videos exceed its payload limits and cannot be stored as documents. Azure Blob Storage is the correct service for unstructured large files because it supports up to ~4.77 TB per blob and is optimized for high-throughput streaming. A common pattern is to store the file in Blob Storage and keep only a URL or metadata reference inside Cosmos DB, using Cosmos for querying metadata rather than for binary content itself.

Go deeper

Related to this question