Network Model

NVIDIA Minerva automatically discovers network components and represents them as a graph of objects and relationships. The graph supports multi-level object hierarchies and peer relationships, and associates metrics, events, and logs derived from time-series and logged data with each object and relationship. Metrics are represented as Measures. Each object has a Type — such as Device, Interface, or Queue — and its properties and relationships are represented as Dimensions.

Dimensions

A Dimension can represent a property of an object (for example, the ifName of an interface), an identifier, or a property of a related object. Dimensions help organize, view, and filter data. For example, a dimension can identify whether an IP address belongs to a management device, a flow source, or a probe agent. NVIDIA Minerva populates dimension information automatically as data is ingested. If a dimension has no applicable value for a given data point, it is labeled Null.

Measures

NVIDIA Minerva provides a library of measures that define the metrics in collected data. When paired with dimensions, measures let you query and filter data using UI visualizations or query APIs.

• Stack measures to view multiple measures together. For example, view ifInDiscards and ifOutDiscards at the same time.

• Build views, pick measures, and filter on measure values. For example, to view CPU usage that exceeds 90% in heatmaps, specify a minimum value of 90 for the Utilization measure.

Data Sources

When data is collected or generated, it is stored as a particular data source type. For example, syslog data is stored in the syslog_v1 data source. Views can be built from a single data source or a combination of data sources.

Last updated: October 21, 2026