Optimize multi-region cost in Azure Cosmos DB

  • Article

APPLIES TO: NoSQL MongoDB Cassandra Gremlin Table

You can add and remove regions to your Azure Cosmos DB account at any time. The throughput that you configure for various Azure Cosmos DB databases and containers is reserved in each region associated with your account. If the throughput provisioned per hour that is the sum of request units per second (RU/s) configured across all the databases and containers for your Azure Cosmos DB account is T and the number of Azure regions associated with your database account is N, then the total provisioned throughput for your Azure Cosmos DB account, for a given hour is equal to T x N RU/s.

Provisioned throughput with single write region and provisioned throughput with multiple writable regions can vary in cost. For more information, see Azure Cosmos DB pricing.

Costs for multiple write regions

In a multi-region writes system, the net available RU/s for write operations increases N times where N is the number of write regions. Unlike single region writes, every region is now writable and supports conflict resolution. From the cost planning point of view, to perform M RU/s worth of writes around China, you need to configure M RU/s at a container or database level. You can then add as many regions as you would like and use them for writes to perform M RU/s worth of around China writes.

Example

Consider that you have a container in a single-region write scenario. That container is provisioned with throughput of 10K RU/s and is storing 0.5 TB of data this month. Now, let�s assume you add another region with the same storage and throughput and you want the ability to write to the containers in both regions from your app.

This example details your new total monthly consumption:

Item Usage (monthly) Rate Monthly Cost
Throughput bill for container in China North (single write region) 10K RU/s * 24 hours * 31 days CNY0.051 per 100 RU/s per hour CNY 3,794.4
Throughput bill for container in 2 regions - China North & China East (multiple write regions) 2 * 10K RU/s * 24 hours * 31 days CNY0.102 per 100 RU/s per hour CNY 15,177.6
Storage bill for container in China North 0.5 TB (or 512 GB) CNY 2.576/GB CNY 1318.92
Storage bill for container in 2 regions - China North & China East 2 * 0.5 TB (or 1,024 GB) CNY 2.576/GB CNY 2637.83

Note

This example assumes 730 hours in a month.

Improve throughput utilization on a per region-basis

If you have inefficient utilization, you can take steps to make the maximum use of the RU/s in read regions by using change feed from the read-region. Or, you can move to another secondary if over-utilized. For example, one or more under-utilized read regions is considered inefficient. You need to ensure you optimize provisioned throughput (RU/s) in the write region first.

Writes cost more than reads for most cases excluding large queries. Maintaining even utilization can be challenging. Overall, monitor the consumed throughput in your regions and add or remove regions on demand to scale your read and write throughput. Make sure to understand the effect to latency for any apps that are deployed in the same region.

Related content