RU estimation in Cosmos DB

Question

RU estimation in Cosmos DB

ElenaMarlowe-3376 85

Hi,

For migrate Cosmos DB from serverless to provisioned mode, since usage can vary over time, I’m trying to figure out the best way to estimate the required RU/s before switching.

I’ve checked the "Monitoring" and "Metrics" section in the portal and can view total Request Units (RUs), but I’m not quite sure what value I should enter into the Azure Pricing Calculator. Should I base it on peak RU/s, or is it better to use an average? And should the unit be considered per second?

Any suggestions or best practices from your experience would be appreciated.

1 answer

Your answer

Answer 1

Chiugo Okpala 1,905 MVP

@ElenaMarlowe-3376 welcome to the Microsoft Q&A community.

N/B: I have generated the below answer using co-pilot as an AI tool. Also I have validated and updated the AI output.

Estimating Request Units (RU/s) when transitioning from serverless to provisioned throughput in Azure Cosmos DB requires careful analysis of usage patterns. Here’s how you can approach it:

Key considerations for RU estimation

Peak vs. Average RU/s:
- Peak RU/s ensures you have enough capacity to handle the highest bursts of activity without throttling. If your workload has significant spikes, provisioning based on peak usage would be safer.
- Average RU/s gives a more cost-effective baseline but may lead to throttling if your workload fluctuates significantly.
- A balanced approach is to consider peak RU/s but apply autoscale to optimize costs while ensuring performance.
Units per second:
- Since Cosmos DB provisioned mode charges based on RU/s, ensure that your calculations are done per second and not per minute/hour.

Best practices for estimating RU/s

Analyze historical data Check the Metrics section in Azure Portal, especially the Total Request Units and Max RU/s consumed over a reasonable period. If your workload varies daily, a week’s worth of data provides a better estimate.
Use Azure Monitor alerts Set up alerts on RU consumption to track unexpected spikes.
Simulate the workload If feasible, create a test environment where you gradually scale RU/s up and down to see what works best.

Thus in Azure Pricing calculator:

If you going to be doing manual provisioning, enter th peak RU/s value
If you going to be using autoscale, enter minimum expected RU/s, and Azure will sacle up when needed (up to 10x your base RU allocation).

I hope these helps. Let me know if you have any further questions or need additional assistance.

Also if these answers your query, do click the "Upvote" and click "Accept the answer" of which might be beneficial to other community members reading this thread.

User's image

Vijayalaxmi Kattimani 3,250 Reputation points Microsoft External Staff Moderator

2025-06-16T03:52:19.05+00:00

Hi ElenaMarlowe-3376

Following up to see, we haven’t heard from you on the last response provided by Chiugo Okpala

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.
ElenaMarlowe-3376 85 Reputation points

2025-06-16T12:34:53.6133333+00:00

Hi @Vijayalaxmi Kattimani ,

I still have a few things I’m not quite clear about. Would it be possible for someone from your team to help explain further? Maybe with an example or any other way that makes it easier to go through?
Vijayalaxmi Kattimani 3,250 Reputation points Microsoft External Staff Moderator

2025-06-17T06:39:44.43+00:00
Hi ElenaMarlowe-3376

Greetings!

We would like to inform you that, when migrating Azure Cosmos DB from serverless to provisioned throughput, estimating the appropriate RU/s (Request Units per second) is key to ensuring performance without overpaying. Since usage varies, here are some best practices and suggestions:

Understand What RU/s Means:

In provisioned mode, you commit to a certain number of RU/s per second, which is reserved 24/7.

In serverless, you pay per RU consumed, but with a cap on burst traffic (~5,000 RU/s per container).

You can Estimate Required RU/s using Metrics from Serverless Usage:

Navigate to: Azure Portal → Cosmos DB Account → Monitoring → Metrics → Total Request Units → Aggregation: Max, Granularity: 1 minute

Look at “Max RU/s per minute” for peak traffic over a representative period (e.g., the past 7 or 30 days). Use the peak RU/s per second from the busiest 1-minute window as a baseline.

You can determine Peak vs Average: Provisioned mode charges for the configured RU/s 24/7, regardless of usage. Peak RU/s ensures performance but may result in higher cost.Average RU/s is cheaper but may result in throttling during spikes.

We would suggest you that, you can use peak or 95th percentile RU/s for production workloads that require consistency. For dev/test or tolerant workloads, average + buffer (e.g., 30%) is fine.

Example 1: a workload is expected to burst to a maximum of 500 RU/s and consume a total of 20,000,000 RUs over a month.

In provisioned throughput mode, you would configure a container with provisioned throughput at a quantity of 500 RU/s for a monthly cost of: $0.008 * 5 * 730 = $29.20

In serverless mode, you would pay for the consumed RUs: $0.25 * 20 = $5.00

Example 2: a workload is expected to burst to a maximum of 500 RU/s and consume a total of 250,000,000 RUs over a month.

In provisioned throughput mode, you would configure a container with provisioned throughput at a quantity of 500 RU/s for a monthly cost of: $0.008 * 5 * 730 = $29.20

In serverless mode, you would pay for the consumed RUs: $0.25 * 250 = $62.50

I would request you to refer this below mentioned documents for more information.

https://learn.microsoft.com/en-us/azure/cosmos-db/throughput-serverless#detailed-comparison

https://learn.microsoft.com/en-us/azure/cosmos-db/nosql/how-to-change-capacity-mode

https://learn.microsoft.com/en-us/azure/cosmos-db/throughput-serverless

https://learn.microsoft.com/en-us/azure/cosmos-db/set-throughput

https://azure.microsoft.com/en-us/pricing/calculator/

I hope this information helps. Please do let us know if you have any further queries.
ElenaMarlowe-3376 85 Reputation points

2025-06-17T13:38:53.5533333+00:00

Hi @Vijayalaxmi Kattimani ,

When I use the pricing calculator and select Autoscale provisioned, there's a field called "Average % Utilization", does that refer to the 10%–100% autoscale range?

Also, where can I find a good reference value to fill in there? Should I look at the "Normalized RU Consumption (avg)" metric in the portal?
Vijayalaxmi Kattimani 3,250 Reputation points Microsoft External Staff Moderator

2025-06-18T04:33:59.8233333+00:00
Hi ElenaMarlowe-3376

Here's what the “Average % Utilization” field means and how to pick the right value:

When you select Autoscale (Provisioned) in the Azure Pricing Calculator:

You enter a maximum RU/s (e.g., 10,000 RU/s).

The “Average % Utilization” field tells the calculator how much of that maximum RU/s you typically consume on average.

Note: Azure Cosmos DB autoscale provisions between 10% and 100% of your configured max RU/s. You are billed for the max RU/s used in any given hour, not for actual average RU usage. The calculator uses average % utilization to estimate your likely bill across time.

How to Pick a Realistic “Average % Utilization”

Go to Metrics in your Cosmos DB instance in the portal and select:

Normalized RU Consumption (avg)

Set the aggregation granularity to 1 hour.

This value is a percentage of the max RU/s being used.

Example:

If you plan to autoscale to 10,000 RU/s, and the “Normalized RU Consumption (avg)” shows 35%, that means you’re using ~3,500 RU/s on average. In this case, enter 35% into the "Average % Utilization" field in the pricing calculator.

I hope this information helps. Please do let us know if you have any further queries.
Vijayalaxmi Kattimani 3,250 Reputation points Microsoft External Staff Moderator

2025-06-19T01:49:45.1333333+00:00

Hi ElenaMarlowe-3376
We haven’t heard from you on the last response and wanted to follow up to check if your issue has been resolved.

If you have found a solution, we would appreciate it if you could share it with the community, as it may be helpful to others. Otherwise, please provide more details, and we will do our best to assist you further.

Share via

RU estimation in Cosmos DB

1 answer

Your answer