Platform Usage

Monitor resource consumption across H2O AI Cloud with the Platform Usage dashboard. Track AI Unit usage and Enterprise h2oGPTe token consumption by workspace, user, service, or model.

The platform collects telemetry data continuously. You can view consumption at hourly intervals for recent date ranges or aggregate data to daily, weekly, monthly, or yearly views for longer time periods.

Prerequisites

To access the Platform Usage dashboard, you must belong to both of these groups:

• admins
• telemetry-admin

To access Platform Usage, navigate to the dropdown on the top right corner, and select Platform Usage:

note

Contact your platform administrator if you need access to these groups.

Accessing Platform Usage

Select Platform Usage from the main navigation menu. The dashboard loads with the default view:

Setting	Default value
Resource	AI Unit Usage
Date range	Last 30 days
Granularity	Daily
Group by	Workspace

Resource types

The dashboard tracks two resource types. Select a resource type from the Resource dropdown in the filter panel.

AI Unit Usage

AI Units are a normalized compute currency that represents CPU, memory, and GPU consumption.

Capability	Options
Group by	Workspace, User, Service, Tag
Metrics	AI Units, CPU (Cores), Memory (GiB), GPU
Filter by	Workspace, User, Service, Tag

Token Usage

Tokens represent large language model (LLM) consumption for input prompts and output completions.

Capability	Options
Group by	Workspace, User, Model
Metrics	Total Tokens, Input Tokens, Output Tokens
Filter by	Workspace, User, Model

To view Token Usage for Enterprise h2oGPTe, select H2OGPTe Token Usage from the dropdown:

Dashboard components

Statistics cards

Summary cards appear at the top of the dashboard.

AI Unit Usage

Card	Description
Max AI Units	Highest single time-period total across all groups
Average AI Units	Total AI Units divided by number of time periods (label changes based on granularity, e.g., "Average Daily AI Units")
Group Count	Number of unique groups (label changes based on grouping, e.g., "Workspace Count" or "User Count")

Token Usage

Card	Description
Total Tokens	Sum of all tokens across the date range
Input Tokens	Sum of prompt tokens
Output Tokens	Sum of completion tokens
Group Count	Number of unique groups

The dashboard displays token values with K (thousands) and M (millions) suffixes for readability.

Stacked bar chart

The main visualization displays consumption over time:

• X-axis: Time periods based on the selected granularity
• Y-axis: Metric values (AI Units, CPU cores, memory GiB, GPU count, or tokens)
• Stacked bars: Each color represents one group; bar height shows the total for that time period

To see detailed information, hover over any bar. The tooltip displays:

• Time period label
• Total value across all groups
• Per-group breakdown with color indicators

A legend below the chart maps each color to its group name. To reload data, select the refresh icon in the chart header.

Data table

The table below the chart presents the same data in tabular format:

• First column: Group names with color indicators matching the chart (sticky when scrolling)
• Remaining columns: One column per time period (scrollable horizontally)
• Values: Formatted based on metric type (decimals for CPU/Memory, integers for GPU, abbreviations for tokens)

CSV export

To export data as CSV, select the download icon above the data table.

AI Unit Usage

Includes all four metrics (AI Units, CPU, Memory, GPU) as separate columns per time period.

Token Usage

Exports the currently selected token type.

File naming: resource_usage_{startDate}_{endDate}_{groupBy}.csv

Filters and controls

All controls are in the filter panel on the right side of the dashboard.

Resource selector

Switch between AI Unit Usage and Enterprise h2oGPTe Token Usage. When you change the resource type:

• The Group By dropdown updates to show resource-specific options
• The metric selector changes to match the resource type
• Entity filters reset to match the new resource type

Date range

To open the date range picker, select the date range button.

Quick presets (auto-fill dates and granularity):

Preset	Date range	Default granularity
1 Day	Yesterday to today	Hourly
7 Days	7 days ago to today	Daily
MTD	First of current month to today	Daily
1 Month	30 days ago to today	Daily
3 Months	90 days ago to today	Daily

Custom date range: Select any start and end date using the date pickers. The end date maximum is today; the start date maximum is yesterday.

Timezone: All dates display in your browser's local timezone.

A summary below the date range button shows the current granularity setting.

To update the dashboard, select Apply. To discard changes, select Cancel.

Granularity

Granularity controls the time bucket size. Available options depend on the selected date range:

Granularity	Minimum range	Constraint	Example label
Hourly	Any	Maximum 100 hourly buckets	"1/15, 3:00 PM"
Daily	2 days	Maximum 100 daily buckets	"1/15/2026"
Weekly	14 days	Maximum 100 weekly buckets	"Jan 15, 2026"
Monthly	60 days	Maximum 100 monthly buckets	"Jan 2026"
Yearly	730 days	No maximum	"2026"

If the current granularity becomes invalid for a new date range, the dashboard auto-corrects to a valid option.

Group by

Select how data is aggregated in the chart and table.

AI Unit Usage

Option	Description
Workspace	Group by workspace name
User	Group by user display name
Service	Group by platform service
Tag	Group by resource tag

Token Usage

Option	Description
Workspace	Group by workspace name
User	Group by user display name
Model	Group by LLM model name

Changing the group updates the chart immediately using prefetched data.

Metric type

AI Unit Usage

Select which metric to display:

Metric	Unit
AI Units	AI Units (normalized compute)
CPU (Cores)	CPU core count
Memory (GiB)	Memory in gibibytes
GPU	GPU count

Changing the metric updates the chart Y-axis label and values without refetching data.

Token Usage

Select which token metric to display:

Token type	Description
Total Tokens	Input plus output tokens
Input Tokens	Prompt tokens only
Output Tokens	Completion tokens only

Entity filters

Multi-select dropdowns let you filter data to specific entities.

AI Unit Usage

Available filters: Workspace, User, Service, Tag

Token Usage

Available filters: Workspace, User, Model

How to filter:

1. Select items in the filter dropdowns. Checkmarks indicate pending selections.
2. Select Apply to fetch filtered data.
3. Select Reset to clear all filters and show unfiltered data.

Button states:

• Apply is enabled when you have filter selections that haven't been applied yet.
• Reset is enabled when any filters are active (pending or applied).

note

When you change the date range, all entity filters automatically reset.

Sharing and bookmarking

Dashboard settings persist in the browser URL as query parameters:

• Copy the URL to share your exact view with colleagues
• Bookmark the URL to save a specific configuration
• Use browser back and forward buttons to navigate through your view history

URL parameters

Parameter	Values	Default
resource	ai_units, h2ogpte_tokens	ai_units
groupBy	workspace, user, service, tag, model	workspace
metric	ai_units, cpu, memory, gpu	ai_units
startDate	YYYY-MM-DD format	30 days ago
endDate	YYYY-MM-DD format	Today
granularity	hourly, daily, weekly, monthly, yearly	daily

Default values are omitted from the URL to keep URLs concise.

Example URLs

Default view (30 days, daily, workspace, AI Units):

/platform-usage

CPU usage by user for January 2026:

/platform-usage?groupBy=user&metric=cpu&startDate=2026-01-01&endDate=2026-01-31&granularity=daily

GPU usage by service:

/platform-usage?groupBy=service&metric=gpu

Enterprise h2oGPTe token usage by model:

/platform-usage?resource=h2ogpte_tokens&groupBy=model

Weekly workspace usage over 3 months:

/platform-usage?startDate=2025-11-01&endDate=2026-02-01&granularity=weekly

Monthly usage for a full year:

/platform-usage?startDate=2025-01-01&endDate=2026-01-01&granularity=monthly

Hourly usage for yesterday:

/platform-usage?startDate=2026-02-09&endDate=2026-02-10&granularity=hourly

Common tasks

Compare two workspaces over 3 months

1. Set the date range to 3 Months with Daily granularity.
2. Set Group By to Workspace.
3. In the Workspace filter, select the two workspaces to compare.
4. Select Apply.

View GPU usage by service for the past week

1. Set the date range to 7 Days with Daily granularity.
2. Set Group By to Service.
3. Set Metric Type to GPU.

Identify which LLM models consume the most tokens

1. Set Resource to Enterprise h2oGPTe Token Usage.
2. Set Group By to Model.
3. Set Token Type to Total Tokens.
4. The chart displays token consumption by model.

Export a monthly CPU usage report

1. Set a custom date range covering the desired months.
2. Set Granularity to Monthly.
3. Set Group By to your preferred dimension.
4. Select the CSV download button above the table. The exported CSV includes all four metrics per time bucket.

Limitations

• Granularity constraints: If the selected granularity becomes invalid for a new date range, the dashboard auto-corrects to a valid option.
• Filter reset: Entity filters automatically reset when you change the date range. Filter state is not persisted in the URL.
• Metric parameter scope: The metric URL parameter only applies to AI Unit Usage. It's ignored when viewing Enterprise h2oGPTe Token Usage.
• Group by compatibility: The service and tag group options are only available for AI Unit Usage. If you switch to Enterprise h2oGPTe Token Usage with these groupings, the dashboard resets to workspace.
• Backward compatibility: The URL parameter groupBy=component maps to service for backward compatibility with older bookmarks.
• Invalid parameters: Unknown or malformed URL parameter values are silently ignored and defaults are used.

---

Feedback

• Submit and view feedback for this page
• Send feedback about AI App Store to cloud-feedback@h2o.ai