Processor Workload Forecasting Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional workload prediction models for processors are inflexible and lack accuracy, leading to inappropriate frequency and voltage generation, which results in inefficient power consumption and resource allocation due to their reliance on fixed forecasting methods that fail to adapt to varying workload histories.
Innovation Solution
A system and method that dynamically selects the most appropriate forecasting model from a pool of models based on historical workload data, using techniques like ARIMA, Weighted Moving Average, or Exponential Smoothing, to predict future workloads with minimal error, thereby optimizing power consumption by generating accurate frequency and voltage settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed forecasting model is used for workload prediction, then the model structure is simple and easy to implement, but the prediction accuracy deteriorates when workload profiles vary
Solution Approach 1:
The patent implements a dynamic model selection mechanism that automatically chooses different forecasting models based on the characteristics of the current workload history. Instead of using a single fixed model, the system evaluates multiple models (ARIMA, Exponential Smoothing, Moving Average) and selects the most appropriate one for each prediction interval, making the system adaptable to varying workload patterns while maintaining manageable complexity through automated selection.
Solution Approach 2:
The patent changes the parameter of model selection from static (fixed model) to dynamic (model chosen based on workload characteristics). By introducing parameters that characterize workload patterns and using these to select from multiple forecasting models with different parameters, the system achieves high accuracy across varying workload profiles without requiring an overly complex unified model.
2Measurement precision
If workload prediction accuracy is improved by using multiple forecasting models, then power consumption optimization is enhanced, but device complexity increases
Solution Approach 1:
The patent implements a self-service mechanism where the system automatically evaluates multiple forecasting models against current workload data and selects the best-performing model without external intervention. The error computation and model selection are performed autonomously by the system itself, reducing the need for complex external control mechanisms while achieving high prediction accuracy through multiple models.
Solution Approach 2:
The patent introduces a feedback loop where prediction errors are computed and used to select the next forecasting model. By continuously monitoring prediction accuracy and adjusting model selection based on feedback from previous predictions, the system achieves high overall accuracy without requiring all models to be equally complex, as only the best-performing model is actively used at any given time.
3Loss of energy
If dynamic voltage and frequency scaling is implemented based on accurate workload prediction, then power consumption is reduced, but the complexity of voltage and frequency control increases
Solution Approach 1:
The patent applies preliminary action by using forecasting models to predict future workload before it actually occurs. This allows the system to proactively adjust voltage and frequency settings in advance based on predicted workload patterns, optimizing power consumption before the workload change happens rather than reacting after the fact, thereby reducing overall power dissipation while managing control complexity through advance planning.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and systems for selecting an appropriate forecasting model for accurate workload prediction of a processor 102 is disclosed. The processor 102 is configured to monitor workload and extract workload history of the processor 102 for a given time interval. Further, the processor is configured to create plurality of forecasting models based on the extracted workload history and apply the group of forecasting model on the extracted workload history to obtain a plurality of predicted future workload for the given time interval. Further, the processor 102 is configured to compute an error measure of the plurality of predicted future workload in reference to an actual workload of the processor 102 and select the appropriate forecasting model from plurality of the forecasting models having least error measure among the computed error measures for dynamically scaling frequency and voltage required by the processor and thereby optimizing energy consumption in the processor.