Skip to main content

🌍 Azure Cosmos DB (Interview Guide)

 

🟢 What is Azure Cosmos DB?

  • A globally distributed, multi-model, NoSQL database service.
  • Designed for low latency, high availability, and global scalability.
  • Supports multiple APIs: SQL (Core), MongoDB, Cassandra, Gremlin (Graph), Table API.

👉 Think of it as a database that automatically scales across the globe.

🟢 Key Concepts

  • Account → Database → Container → Items
  • Partition Key → spreads data evenly across physical partitions.
  • Request Units (RUs) → performance currency (every operation consumes RUs).
  • Consistency Levels → trade-off between performance and accuracy:
    • Strong, Bounded Staleness, Session, Consistent Prefix, Eventual.
  • Change Feed → stream of inserts/updates in order.

🟢 Partition Key (Most Asked in Interviews)

  • Good partition key: high cardinality, evenly distributes load.
  • Bad partition key: low cardinality (e.g., true/false, country code).
  • Can use synthetic keys (e.g., CustomerId + Date).

👉 Interviewers love to ask: “What happens if you choose a bad partition key?”
Answer: It creates hot partitions → performance issues.

🟢 Pricing Models

  • Provisioned Throughput (RU/s) → fixed capacity, predictable cost.
  • Autoscale → scales RU/s up/down automatically.
  • Serverless → pay per request (for small workloads).

🟢 Example: Insert Document in .NET

using Azure.Cosmos;

string endpoint = "<COSMOS_ENDPOINT>";
string key = "<COSMOS_KEY>";
CosmosClient client = new CosmosClient(endpoint, key);

var db = await client.CreateDatabaseIfNotExistsAsync("SampleDb");
var container = await db.CreateContainerIfNotExistsAsync("Products", "/category", 400);

var item = new { id = Guid.NewGuid().ToString(), category = "electronics", name = "Laptop", price = 1200 };
await container.CreateItemAsync(item, new PartitionKey(item.category));

Console.WriteLine("✅ Item inserted!");

🟢 Example: Query Documents (SQL-like)

var query = "SELECT * FROM c WHERE c.price > 500";
var iterator = container.GetItemQueryIterator<dynamic>(query);

while (iterator.HasMoreResults)
{
    foreach (var doc in await iterator.ReadNextAsync())
        Console.WriteLine($"{doc.name} - {doc.price}");
}

🟢 Advanced Features

  • Transactional Batch → atomic operations within same partition.
  • Stored Procedures (JS) → run logic close to data.
  • Change Feed → react to changes in real time (e.g., with Azure Functions).

🟢 Monitoring

  • Integrated with Azure Monitor + App Insights.
  • Track RU consumption, latency, throttling (429 errors).
  • Use retry policies with exponential backoff.

🟢 Common Interview Questions

  1. What is Azure Cosmos DB?
  2. What is a partition key? How do you choose one?
  3. What are Request Units (RUs)?
  4. Explain different consistency levels.
  5. What is the Change Feed? Use case?
  6. Difference between Provisioned Throughput and Serverless mode?
  7. How does Cosmos DB achieve global distribution?
  8. What are hot partitions and how do you avoid them?
  9. Difference between Cosmos DB and SQL Database?
  10. How do you monitor performance and troubleshoot 429 errors?

🟢 Pros & Cons

✅ Pros

  • Global distribution, low latency.
  • Multi-model, flexible APIs.
  • Tunable consistency.
  • Change Feed support for event-driven apps.

❌ Cons

  • Cost can grow with poor partition design.
  • Limited support for complex joins (compared to SQL).
  • RU model can be confusing for beginners.

✅ Quick Recap

  • Cosmos DB = globally distributed NoSQL database.
  • Partition Key → most critical design choice.
  • RU/s → performance unit (watch consumption).
  • Consistency Levels → balance between latency & accuracy.
  • Change Feed → real-time stream of changes.


Comments

Post a Comment