Power Predictor Circuitry for Server Workload Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing power prediction models for data centers face inaccuracies due to variations in compute device designs and environmental factors, leading to inconsistent power consumption estimates across different instances of the same product, and limitations in model types used, which struggle with predicting power consumption across the full range of workloads.
Innovation Solution
The development of a power predictor circuitry that divides the range of workloads into sub-ranges and generates specific prediction models for each sub-range using both historical and extrapolated data, selecting the model with the lowest error to provide accurate power consumption estimates for each compute device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single power prediction model is used for all workloads, then the model structure is simple, but the prediction accuracy deteriorates across different workload ranges
Solution Approach 1:
The patent divides the workload range into multiple sub-ranges and creates separate prediction models for each sub-range. This segmentation allows each model to be specialized for specific workload conditions, improving prediction accuracy without requiring an excessively complex unified model structure.
2Ease of manufacture
If prediction models are trained only on historical data, then the training process is simple, but the prediction reliability deteriorates when historical data is insufficient
Solution Approach 1:
The patent combines historical workload data with extrapolated workload data to form a comprehensive training dataset. This merging of data sources ensures sufficient training data availability while maintaining the simplicity of the training process, thereby improving prediction reliability.
3Ease of operation
If limited model types are used for power prediction, then the implementation is straightforward, but the adaptability deteriorates across diverse workload scenarios
Solution Approach 1:
The patent employs multiple model types with different parameter configurations to handle diverse workload scenarios. By changing model parameters and selecting appropriate models based on workload characteristics, the system achieves high adaptability while maintaining straightforward implementation through systematic model selection.
Data Source
AI summary
Methods, apparatus, systems, and articles of manufacture are disclosed to predict power consumption in a server. An example apparatus includes interface circuitry to obtain a power prediction request corresponding to the server range determiner circuitry to divide a training data set into a first sub-range of data and a second sub-range of the data; a data point in the training data set representative of resource utilization of a workload and a corresponding power consumption metric of the workload; model trainer circuitry to train first candidate models based on the first sub-range of the data and second candidate models based on the second sub-range of the data; and prediction selector circuitry to: select a first prediction model from the first candidate models; and select a second prediction model from the second candidate models, outputs of the first and the second prediction models to predict the power consumption of the server.