KodeKloud Notes

In this guide, you’ll learn how to create and optimize an Azure Cosmos DB account for your global, low-latency applications. We’ll cover account creation, storage scaling, partition key design, consistency tuning, and throughput provisioning.

1. Create a Cosmos DB Account

Sign in to the Azure portal.
Go to Azure Cosmos DB and click + Create.
Enter a unique account name, then select your subscription, resource group, and preferred region.
Choose the API for your workload. Options include:
- Core (SQL)
- MongoDB
- Cassandra
- Gremlin
- Table
- Azure Cosmos DB for PostgreSQL
API Selection
You can select only one API per account. For JSON-like documents or MongoDB migrations, choose MongoDB.
(Optional) Enable Multi-region writes to allow write operations in all selected regions.
Add additional regions for global distribution and high availability.

2. Scale Storage with Containers & Partitions

Azure Cosmos DB distributes your data across containers and partitions for virtually unlimited storage and throughput:

Container: A logical namespace for your items (documents, rows, edges). Use multiple containers to isolate workloads.
Partition: Each container is automatically sharded into physical partitions up to 20 GB. Proper partition key design ensures balanced data distribution.

The image illustrates the capacity structure of Cosmos DB, showing how containers are divided into partitions, with each partition having a size limit of 20 GB. It also highlights the ability to add more storage by adding containers.

Concept	Description
Container	Namespace for items; scales throughput at container or database level
Partition	Physical shard up to 20 GB; automatically managed by Cosmos DB

3. Design Effective Partition Keys

A well-chosen partition key is crucial for performance and cost optimization:

Use a property that is frequently present in your queries.
Ensure high cardinality and even distribution to avoid “hot” partitions.

Partition Key Best Practices

Select a key with many unique values (e.g., userId or orderId) to spread traffic and storage evenly.

The image illustrates the concept of partition keys in Cosmos DB, showing how partitions are used to speed up searching and ensure even distribution without hotspots. It includes a diagram with two containers, each containing different regions.

4. Configure Consistency Levels

Azure Cosmos DB provides five consistency levels tailored to different SLAs:

Level	Latency	Failure Impact	Guarantee
Strong	High	Higher failure risk	Reads always see the latest committed write
Bounded Staleness	Medium	Moderate	Reads lag by a fixed version interval or time window
Session	Low	Low	Monotonic reads/writes within a session
Consistent Prefix	Low	Low	Reads never see out-of-order writes
Eventual	Lowest	Lowest	Reads may be stale until replicas catch up

The image illustrates a world map with icons representing data centers or regions, highlighting the flexibility of Cosmos DB in providing complete and eventual consistency. It emphasizes that no write is complete until every region is updated.

Choose the level that balances latency, throughput, and your application’s tolerance for stale reads.

5. Provision Throughput & Manage Request Units (RUs)

Throughput in Cosmos DB is expressed in Request Units per second (RUs/sec):

If operations exceed your provisioned RUs, Cosmos DB returns HTTP 429 (throttled) errors.
Throttled requests are retried automatically based on your retry policy, which can increase response times.

Avoid Throttling

Monitor RU consumption in Azure Monitor. Adjust your retry policy and increase RUs if you observe frequent HTTP 429 errors.

The image shows a world map highlighting Mumbai with a focus on transaction throughput and a note about exceeding RU limits leading to throttling.

Estimating & Allocating RUs

Profile your workload: reads/writes per second and item sizes.
Use the Azure Cosmos DB capacity calculator for accurate RU estimates.
Provision at:
- Database level – shared across all containers
- Container level – dedicated throughput per container

Container Name	Provisioned RUs	Use Case
`shipping-orders`	800	High-traffic transactions
`customer-profiles`	100	Low-volume user lookups
`inventory`	100	Periodic stock updates

Links and References

Watch Video

Watch video content