Energy Usage Measurement for Multi-Core Processors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current energy usage monitoring in cloud computing environments, such as data centers and Edge devices, lacks precision at the thread, process, or service level, making it difficult to effectively reduce carbon footprint and energy consumption, as existing methods are often hardware component-based and not sensitive to processor variations.
Innovation Solution
Implementing a system that monitors energy utilization at the thread, process, or service level by tracking microoperations and microarchitectural events, using unique weights to estimate energy usage independently of processor voltage, frequency, and temperature, and providing this data to a telemetry aggregator for accurate energy usage calculation and reporting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If energy usage monitoring is implemented at hardware component level, then measurement coverage is achieved, but measurement precision at thread/process/service level deteriorates
Solution Approach 1:
The patent segments energy monitoring into two layers: hardware component level (processors, memory, storage, network interfaces) and software abstraction level (threads, processes, services). Telemetry aggregators collect data from hardware sensors and aggregate it at software abstraction levels, enabling precise energy measurement at thread/process/service level without requiring direct hardware modification for each software entity.
Solution Approach 2:
The patent introduces telemetry aggregators as intermediary components that collect telemetry data from hardware sensors and process it into software-centric energy metrics. These aggregators act as mediators between the hardware monitoring infrastructure and software energy management systems, translating hardware-level data into meaningful software-level energy consumption information.
2Adaptability or versatility
If processor-specific energy monitoring is used, then measurement accuracy for that processor is improved, but adaptability to processor variations deteriorates
Solution Approach 1:
The patent creates a universal energy monitoring framework that works across different processor architectures, vendors, and models. The system uses standardized telemetry data collection and aggregation mechanisms that can be applied universally to any hardware component, while the software-centric energy metrics provide consistent measurement regardless of underlying processor variations.
Solution Approach 2:
The patent transforms energy measurement from processor-specific physical parameters to software-abstraction-level logical parameters. By measuring energy consumption at the thread, process, or service level rather than at the processor hardware level, the system changes the measurement parameter from hardware-dependent to software-independent, enabling adaptability across processor variations while maintaining measurement accuracy.
Data Source
AI summary
Examples described herein relate to at least one multi-core processor and a circuitry can determine and output energy usage of a process regardless of a core of the at least one multi-core processor that executes the process. The circuitry can determine the energy usage of the process based on cache operations and processor microoperations associated with the process. The energy usage of the process can be based on dynamic capacitance (Cdyn) levels and one or more of: temperature of the at least one multi-core processor, input voltage temperature to the at least one multi-core processor, and/or frequency of the at least one multi-core processor.