AI Units
An AI Unit measures maximum resource consumption in the H2O AI Cloud (HAIC). A formal definition of an AI Unit is as follows:
An AI Unit is an H2O AI Cloud (HAIC) unit of measurement that measures the consumption of CPUs, GBs (gigabytes of RAM), and GPUs across its platform for a particular organization.
- AI Unit Consumption provides a graph illustrating an organization's up-to-date consumption of AI Units in the H2O AI Cloud (HAIC). In particular, it illustrates the maximum capacities of AI Units used over time in HAIC. See Graph: AI Units to learn more about the graph.
- AI Unit Consumption uses the AI Unit Consumption (AIUC) formula to determine the maximum capacities of AI Units used over time in HAIC. To learn more about AIUC, see AI Units calculation.
Graph: AI Units
To view the maximum capacities of AI Units used over time in HAIC, AI Units Consumption offers the following graph:
- X-axis
- The X-axis represents time, beginning with the first time an organization started using the H2O AI Cloud (HAIC). The data display through the X-axis starts in an hourly form, but AI Unit Consumption aggregates it up as needed to improve the data visualization. For example:
- If there are more than seven days of data, the graph will display days instead of hours across its X-axis.
- If there are more than four weeks of data, the graph will display weeks instead of hours across its X-axis.
- If there are more than 12 months of data, the graph will display months instead of hours across its X-axis.
- The X-axis represents time, beginning with the first time an organization started using the H2O AI Cloud (HAIC). The data display through the X-axis starts in an hourly form, but AI Unit Consumption aggregates it up as needed to improve the data visualization. For example:
- Y-axis
- The Y-axis represents the number of AI Units consumed.
- AI Units encapsulate resource consumption and, therefore, HAIC charge peak usage of resources (e.g., GPU consumption). For example, if you use 10 AI Units for 10 days, HAIC will charge you for 10 AI Units and not 100 AI Units (10 AI Units x 10 days). The AI Units graph does not display a total number of AI Units consumed but rather the number of AI Units consumed at different moments (over time).
AI Units calculation
To calculate the maximum capacity of AI Units consumed at a given point, AI Unit Consumption uses the following formula:
- GPU: Measured in GPU cores (no distinction in GPU types is made)
- CPU: Measured in K8s vCPU units
- RAM: Measured in GBs (gigabytes)
AI Unit Consumption counts resource requests (not used) of CPUs, GBs of RAM, and GPUs as AI Units are based on peak aggregate usage of HAIC.
Imagine that organization X has already consumed 1 GPU, 32 vCPUs, and 64 GBs of RAM since its first use of HAIC. Using the formula above, we can say that X at this point has consumed 7 AI Units since its first use of HAIC.
- Submit and view feedback for this page
- Send feedback about AI Unit Consumption to cloud-feedback@h2o.ai