Metering and usage experience

Runtime-Level metering

Metering is now performed at the runtime level, capturing start, stop, and pause events for each runtime tied to an engine. Presto Engines: One-to-one mapping (single runtime per engine). Spark Engines: Multiple runtimes per engine, including subtypes like:

• Kernel
• HistoryServer
• Application
• Each Spark application creates a new runtime, and active/inactive hours are tracked individually.

Features for metering

Runtime activity bars display active and inactive hours per runtime. The View History links are scoped to the runtime level for detailed event tracking. The UI captures start and end timestamps for every engine session (Presto++, Spark, Milvus, Gluten). Record engine type and T-shirt size (S/M/L) for pricing alignment. Provide hourly RU usage data for billing reference.

Reporting considerations

Total usage: It is the sum of individual usage of all the infrastructure components running in the instance for the current month. Total hourly usage: . Usage history per engine: .

Viewing the usage details

To view current billing and usage details, complete the following steps:

1. Log in to the watsonx.data console.

2. From the navigation menu, select Billing and usage. The Billing and usage page opens. You can view the following:

   • Total usage

   It displays the aggregate RU consumption for the current calendar month. Usage Bar:

   • Active RUs (Green): Shows RUs consumed during active engine runtime.

   • Idle RUs (Yellow): Indicates RUs consumed while engines were idle.

   Example: 9.75 RUs total, with 7 RUs active and 2.75 RUs idle.

   • Billable Components Table

   Lists all engines and services contributing to usage.

   Billing and usage

   Field	Description
   Display name	The name of the billable component.
   Type	Engine type (Presto, Spark, Milvus, etc.).
   Size.	T-shirt size (Small, Medium, Large).
   Status	The status of the component. It can be running, failed, provisioning, restarting, resuming, scaling, updating, or paused. The Running status displays a run rate above 0 RUs/hour.
   Total Usage (RUs)	Cumulative RU consumption.
   Total Uptime	Duration the component was active.
   Owner.	Associated user or team.

   When you select an engine (e.g., Presto Java), you can view the detailed metrics:

   Usage: Active RUs vs Idle RUs breakdown.

   Runtime:

   Uptime and Downtime durations displayed with color-coded bars. Example: Uptime: 1 hour 10 min, Downtime: 1 hour.

   • Total Hourly Usage

   Visual representation of RU consumption across hours and days.

   Color Intensity: Darker shades indicate higher usage.

   Helps identify peak usage periods for optimization.

   • Usage History per Engine

   Line chart showing RU usage trends for individual engines over time. Multiple Engines: Each engine represented by a different color line. Enables comparison of usage patterns across engines.

Metering and billing granularity in watsonx.data

Billing is based on the the total run rate, which is the sum of the individual run rates of all the infrastructure components currently running in your instance. Resource unit consumption begins as soon as Support Services are activated with the initiation of Quick Start step and continues until the instance is deleted.

You can view the estimate of the expected per-hour run rate consumption of Resource Units from the About tab in the watsonx.data IBM Cloud® catalog page.

Important: watsonx.data is metered on a per-second basis to ensure precise and granular tracking of usage.

This fine-grained metering allows for greater transparency and control, capturing every second the instance is active, regardless of whether any engine is enabled. While Usage and Billing collection is aggregated and updated on IBM Cloud at an hourly basis, the underlying metering collection is done at a per second basis. This continuous per-second metering ensures that charges reflect actual usage time. To avoid incurring unnecessary costs, it is important to delete unused instances promptly.

For example:

Scenario 1: Presto C++ starter instance usage

You create a watsonx.data instance configured with Presto C++ starter, which includes:

Presto C++ starter, and support services enabled.

The following table displays the metering calculation when you run the instance for 14 minutes 30 seconds. The usage time which, is monitored in seconds will be internally converted to hours. Here, 0.2416 hours.

Scenario 2: Instance with Presto C++ starter and Milvus starter

You create a watsonx.data instance with the following configuration:

Presto C++ starter, Milvus starter size and support services enabled.

The following table displays the metering calculation when you run the instance for a total of 35 minutes and 30 seconds but Presto C++ runs for only the first 14 minutes and 30 seconds.