Azure Cosmos DB

Learning objectives#

• Model data as JSON documents
• Choose a good partition key
• Understand request units and consistency
• Create and query items with the .NET SDK

Explanation#

Data lives in accounts → databases → containers → items (JSON documents). Different items in a container can have different shapes.

Every container has a partition key. Cosmos DB spreads data across physical partitions by this key, so choose one with many distinct values that queries usually include (for example customerId).

Throughput is measured in request units (RU/s). A 1 KB point read costs about 1 RU. Choose provisioned, autoscale or serverless capacity.

Five consistency levels — Strong, Bounded staleness, Session (default), Consistent prefix, Eventual — trade consistency for latency and availability.

Code example: Saving and reading orders#

1using Azure.Identity;
2using Microsoft.Azure.Cosmos;
3

4var client = new CosmosClient("https://cosmos-shop.documents.azure.com:443/", new DefaultAzureCredential());
5Container orders = client.GetContainer("shop", "orders");   // partition key: /customerId
6

7var order = new Order("o-101", "c-1", 56400m, new[] { "Laptop", "Mouse" });
8await orders.CreateItemAsync(order, new PartitionKey(order.customerId));
9

10// Point read: id + partition key — the cheapest operation (~1 RU)
11var read = await orders.ReadItemAsync<Order>("o-101", new PartitionKey("c-1"));
12Console.WriteLine(
quot;{read.Resource.id}: ₹{read.Resource.total} ({read.RequestCharge} RU)");
13

14public record Order(string id, string customerId, decimal total, string[] items);

o-101: ₹56400 (1 RU)

This example is illustrative. Copy or download it and run it in your own environment; the expected output is shown above.

Step-by-step explanation#

1. The client authenticates with Entra ID; no keys.
2. Every item needs an id and its partition key value.
3. A point read uses both, so Cosmos DB goes straight to the right partition.
4. RequestCharge shows the RU cost of each operation — useful for tuning.

Real-world use case#

Common mistakes#

Best practices#

• Pick the partition key from your most common queries
• Prefer point reads (id + partition key)
• Use Session consistency unless you need stronger
• Watch RU charges and use autoscale

Remember this#

Practice exercise#

1/customerId — many distinct values and it matches the common query “orders for this customer”.

Mini quiz#

Want more? Practise all Microsoft Azure MCQs · Microsoft Azure interview questions

Related topics#

My notes#

Saved in this browser.