Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data-driven Software-based Power Estimation for Embedded Devices

Haoyu Wang, Xinyi Li, Ti Zhou, Man Lin

TL;DR

The paper tackles the lack of software-based power measurement for embedded IoT devices and the unreliability of consumer-grade meters for instantaneous readings. It introduces a data-driven pipeline that collects training data from CPU-heavy workloads while recording frequency and utilization traces, then trains models—including a per-frequency polynomial regression—that can predict power with high accuracy. Among the evaluated models, the per-frequency regression achieves $R^2=0.9221$ and $MSE=0.0182$, corresponding to about 92% explained variance, validating the approach on Jetson Nano and Raspberry Pi. A platform-independent kernel module, DataTracker, enables real-time trace collection, making the method practical for energy-aware scheduling and control in devices lacking built-in power measurement capabilities.

Abstract

Energy measurement of computer devices, which are widely used in the Internet of Things (IoT), is an important yet challenging task. Most of these IoT devices lack ready-to-use hardware or software for power measurement. In this paper, we propose an easy-to-use approach to derive a software-based energy estimation model with external low-end power meters based on data-driven analysis. Our solution is demonstrated with a Jetson Nano board and Ruideng UM25C USB power meter. Various machine learning methods combined with our smart data collection & profiling method and physical measurement are explored. Periodic Long-duration measurements are utilized in the experiments to derive and validate power models, allowing more accurate power readings from the low-end power meter. Benchmarks were used to evaluate the derived software-power model for the Jetson Nano board and Raspberry Pi. The results show that 92\% accuracy can be achieved by the software-based power estimation compared to measurement. A kernel module that can collect running traces of utilization and frequencies needed is developed, together with the power model derived, for power prediction for programs running in a real environment. Our cost-effective method facilitates accurate instantaneous power estimation, which low-end power meters cannot directly provide.

Data-driven Software-based Power Estimation for Embedded Devices

TL;DR

The paper tackles the lack of software-based power measurement for embedded IoT devices and the unreliability of consumer-grade meters for instantaneous readings. It introduces a data-driven pipeline that collects training data from CPU-heavy workloads while recording frequency and utilization traces, then trains models—including a per-frequency polynomial regression—that can predict power with high accuracy. Among the evaluated models, the per-frequency regression achieves and , corresponding to about 92% explained variance, validating the approach on Jetson Nano and Raspberry Pi. A platform-independent kernel module, DataTracker, enables real-time trace collection, making the method practical for energy-aware scheduling and control in devices lacking built-in power measurement capabilities.

Abstract

Energy measurement of computer devices, which are widely used in the Internet of Things (IoT), is an important yet challenging task. Most of these IoT devices lack ready-to-use hardware or software for power measurement. In this paper, we propose an easy-to-use approach to derive a software-based energy estimation model with external low-end power meters based on data-driven analysis. Our solution is demonstrated with a Jetson Nano board and Ruideng UM25C USB power meter. Various machine learning methods combined with our smart data collection & profiling method and physical measurement are explored. Periodic Long-duration measurements are utilized in the experiments to derive and validate power models, allowing more accurate power readings from the low-end power meter. Benchmarks were used to evaluate the derived software-power model for the Jetson Nano board and Raspberry Pi. The results show that 92\% accuracy can be achieved by the software-based power estimation compared to measurement. A kernel module that can collect running traces of utilization and frequencies needed is developed, together with the power model derived, for power prediction for programs running in a real environment. Our cost-effective method facilitates accurate instantaneous power estimation, which low-end power meters cannot directly provide.
Paper Structure (32 sections, 2 equations, 9 figures, 6 tables, 2 algorithms)

This paper contains 32 sections, 2 equations, 9 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Limitations of using consumer-grade power meters for energy measurement
  3. Proposed approach: data-driven power estimation
  4. Related works
  5. Data-driven Power Estimation with Consumer-grade Meters
  6. Training Data Collection
  7. Derive Power Model from Collected Data
  8. How to Use the Software-based Power Model?
  9. Building Power Model
  10. Training Data Collection
  11. Time for Running Workload
  12. Frequency Control
  13. Utilization Control
  14. Training Data Illustration
  15. Testing Data Collection
  16. ...and 17 more sections

Figures (9)

  • Figure 1: Power measuring using a consumer-grade meter.
  • Figure 2: An illustration of the methodology
  • Figure 3: Measurement results of single-threaded (top) and multi-threaded (bottom) benchmark
  • Figure 4: Measurement results of 10 benchmarks in Mibench
  • Figure 5: $stringsearch$ vs $bitcount$
  • ...and 4 more figures