YANG Data Model for Energy Measurements in Streaming Devices

The increasing focus on environmental sustainability in streaming media delivery requires accurate measurement and analysis of energy consumption across the delivery chain. This document defines a YANG data model for reporting energy measurements from streaming-capable devices, along with relevant streaming metrics needed for correlation and analysis. The model is designed to support both A-side (mains power) and B-side (internal component) power measurements while maintaining clear distinction between these measurement types. The internal component usage may be used by a collecting entity to estimate the total device power used if such power measurements are not available within the device.

The proposed YANG model defines a comprehensive framework for representing energy measurements in streaming devices. The model is structured around four key components that together provide a complete picture of device energy consumption and its relationship to streaming activities.

The core of the model focuses on both instantaneous and historical power measurements. This dual approach allows for real-time monitoring as well as trend analysis. The model distinguishes between total system power (A-side) and component-level power (B-side) measurements to support different measurement capabilities across device types.

Proper device identification and capability reporting is important for accurate data interpretation. This component allows devices to clearly communicate their measurement capabilities, ensuring that data consumers can properly interpret and validate the reported measurements.

To enable meaningful energy efficiency analysis, the model includes streaming-specific metrics that can be correlated with power measurements. These metrics provide essential context for understanding energy consumption patterns in relation to actual streaming activities.

The following shows the JSON schema for data exchange:

device-id: Opaque identifier for the device. Device identification is important for proper data attribution and analysis.
device-type: The device-type enumeration is chosen to cover the major categories of streaming infrastructure while remaining extensible for future devices. Each device type may have different measurement capabilities and characteristics that affect how their energy data should be interpreted. One of: encoder, decoder, cdn-cache, router, switch, settop-box, tv, mobile-device

The capability reporting mechanism allows devices to explicitly declare their measurement abilities. This is essential because:

Not all devices can measure both mains and component power
Measurement methods may vary in accuracy and reliability
Some devices may only support subset of measurements

This information helps data consumers properly interpret and validate the reported measurements. The following attributes indicates which types of power measurements the device supports:

supports-mains-power: Device can report total power draw
supports-component-power: Device can report component-level power
supports-component-usage: Device can report component-level usage

Power readings are structured to support both high-level and detailed energy analysis:

timestamp: UTC timestamp of measurement. Timestamp precision is critical for correlation analysis
mains-power: Total power draw from mains (A-side measurement). Mains power provides system-level energy perspective
component-power: Individual component power measurements (B-side). Component power enables detailed efficiency analysis
component-usage: Individual component measurements usage (B-side). Component usage enables estimating the power used if such power measurements are not available within the device.

The streaming metrics are selected to enable use-case that require correlation analyses:

active-streams: Number of active streaming sessions as an indicator for system load
total-bandwidth-mbps: Current bandwidth utilization as an indicator for network utilization
video-brightness: Average screen brightness (0-255)
encoding-parameters: Active encoding settings that may affect processing requirements
network-utilization: Network interface utilization (%)

These metrics were chosen based on empirical research showing their relevance to energy consumption patterns.

This draft has been elaborated in the context of the working group 7 of Greening of Streaming.