Cloud Resource Prediction With Asymmetric Loss for SLA Violations

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing cloud resource design systems fail to adjust resource design results to mitigate user requirement violations, leading to potential penalties and profit losses despite heavy penalties for violations.

Innovation Solution

A resource design apparatus that includes a requirement specifying functional unit, a prediction unit, and a model generation unit, which uses machine learning to predict performance and adjust resource settings using a modified loss function to minimize user requirement violations by incorporating a penalty for underperformance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If existing machine learning models with standard loss functions are used for resource design prediction, then prediction accuracy is maintained, but user requirement violation risk cannot be adjusted or reduced

Engineering Contradiction:
Improveuser requirement satisfactionVSAvoidmodel complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent modifies the loss function parameters by adding a penalty term that takes finite values when actual performance falls below predicted performance. This parameter change in the objective function allows the model to prioritize avoiding user requirement violations while maintaining prediction accuracy, directly resolving the contradiction between reliability and model complexity.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces feedback mechanisms where the loss function continuously adjusts predictions based on performance deviations. The penalty term provides feedback when actual performance is lower than predicted, guiding the model to reduce violation risks. This feedback loop enables dynamic adjustment of prediction outcomes to meet user requirements without substantially increasing model complexity.

Inventive Principle:
Principle #23Feedback

2Reliability

If resource design results are adjusted to reduce violation risk, then user requirement satisfaction improves, but prediction accuracy may deteriorate

Engineering Contradiction:
Improveuser requirement satisfactionVSAvoidperformance prediction accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent changes the loss function parameters to include asymmetric penalty terms that penalize underperformance more heavily than overperformance. This parameter modification allows the model to adjust predictions toward higher reliability without sacrificing overall prediction accuracy, as the loss function guides the model to prioritize avoiding violations while maintaining precise performance estimation.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent applies local quality by making the loss function sensitive to specific performance regions where user requirements are violated. The penalty term is activated only when actual performance falls below predicted performance, creating localized adjustment pressure in the underperformance region while maintaining accurate predictions in other regions. This resolves the contradiction by improving reliability locally without deteriorating overall prediction accuracy.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4075273B1Device for preventing cloud service user requirement violation, user requirement violation preventing method, and program
Publication Date: 2025.11.12 NT T INC
  • EP4075273B1 patent drawingFigure 1~2
  • EP4075273B1 patent drawingFigure 3
  • EP4075273B1 patent drawingFigure 4~5

AI summary

A ratio of prediction liable to result in user requirement violation is reduced by adjusting results of resource design even if it is highly likely that the user requirement violation will incur a heavy penalty. There is provided a requirement specifying functional unit (11) that specifies a user requirement for a service of interest, and a resource design unit (12) that predicts, by machine learning, performance achievable at a plurality of resource settings in performing the service of interest and selects a resource setting that satisfies the specified user requirement, based on results of the prediction, wherein the resource design unit (12) generates a P model as a model for use to predict performance, the P model using a P-mode loss function obtained by adding a function to an N model that uses an existing N-mode loss function, the added function taking a finite value when actual performance is lower than predicted performance.