Skip to main content

Usage

There are multiple forms of billable Dremio usage within an organization:

  • Dremio Consumption Units (DCUs) represent the usage of Dremio engines. DCUs are only consumed when your engines are running.
  • Large-language model (LLM) tokens are billed when you use Dremio's AI features via the Dremio-Provided LLM.
  • Storage usage is billed in terabyte-months and only applies to projects that use Dremio-hosted storage. If your projects use an object storage bucket in your account with a cloud provider as the catalog store, storage fees do not apply.

How DCUs are Calculated​

The number of DCUs consumed by an engine depends on two factors:

  • The size of the engine
  • How long the engine and its replicas have been running for

DCU consumption for an engine is calculated as (Total uptime for the engine and its replicas) * (DCU consumption rate for that engine size).

Uptime is measured in seconds and has a 60-second minimum.

The DCU consumption rate for each engine size supported in Dremio is listed in Manage Engines.

DCU Examples​

Example 1​

An organization has two Dremio Cloud engines defined: Engine A and Engine B, where Engine A is a 2XSmall engine, and Engine B is a Medium engine.

Suppose that between 8 a.m. and 9 a.m. one day:

  • Engine A had 2 replicas running for 40 minutes each, so it accumulates a total of 80 minutes of engine uptime.
  • Engine B had 5 replicas running for 50 minutes each, so it accumulates a total of 250 minutes of engine uptime.

The total usage for Engine A for this hour is (80/60) * (16 DCUs/hour) = 21.33 DCUs.

The total usage for Engine B for this hour is (250/60) * (128 DCUs/hour) = 533.33 DCUs.

Example 2​

An organization has one Dremio Cloud engine defined: Engine A, where Engine A is a Medium engine.

Suppose that between 8 a.m. and 9 a.m. one day:

  • Engine A had 1 replica running for the entire hour (60 minutes).
  • Engine A needed to spin up an additional replica for 30 minutes to tackle a workload spike.

Engine A accumulated a total of 90 minutes of engine uptime, so the total usage for Engine A for this hour is (90/60) * (128 DCUs/hour) = 192 DCUs.

How AI Usage Is Calculated​

If you use the Dremio-Provided LLM, you pay directly for the cost of both the input and output tokens used. If you connect to another LLM via your own model provider, you are not currently charged for this usage.

AI Examples​

Example 1​

Say that you use an external model provider as well as the Dremio-Provided LLM to use Dremio's AI features, resulting in a usage footprint like the below:

  • External model provider 500K input tokens used.
  • External model provider: 30K output tokens used.
  • Dremio-Provided: 200K input tokens used.
  • Dremio-Provided: 20K output tokens used.

You are not charged for using Dremio's AI features via an external model. Instead, you are only charged for the tokens consumed by the Dremio-Provided LLM:

  • (200K input tokens)*($1.25/1 million tokens) = $0.25
  • (20K output tokens)*($10.00/1 million tokens) = $0.20

In this scenario, you would be billed for $0.45 of AI feature usage.

In order to simplify the billing experience for AI features Dremio may explore the addition of an AI specific credit, similar to DCUs, in the future.

How Storage Usage Is Calculated​

Storage is calculated through the collection of periodic snapshots of the Dremio-hosted bucket. These snapshots throughout a billing period are averaged through the billing period to calculate a number of billable terabyte-months.

Storage Usage Examples​

Example 1​

Suppose an organization has one Dremio project in a region where the price of a terabyte-month is $23.00, and that in a given month this project:

  • Stores 1 terabyte of data for the entire 30 days of the billing period

Then the total amount charged for the storage would be (1) * ($23.00) = $23.00

Example 2​

Suppose your organization has a project in a region where the price of a terabyte-month is $23.00, and that in a given period this project:

  • Stores 1 terabyte of data for the first 15 days of the month
  • Stores 2 terabyte of data for the last 15 days of the month

On average throughout the month, the project was storing 1.5Tb of data. So the bill would be (1.5) * ($23.00) = $34.5.