Dynamic Resource Prediction in Virtualized Systems

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing resource prediction approaches in virtualized systems are inadequate due to their application-specific nature, lack of adaptability to behavioral changes, and excessive resource allocation, leading to inefficiencies and increased costs.

Innovation Solution

A prediction algorithm that uses Genetic Algorithm (GA) to dynamically determine the optimal size of a sliding window and the optimal number of predicted data, combined with the Kriging method for machine learning-based prediction, and adjusts resource demand based on estimated probability of prediction errors and variable padding, enabling real-time and adaptive resource management.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If over-provisioning is used to meet service level agreement requirements, then reliability of service delivery is improved, but resource utilization deteriorates and energy waste increases

Engineering Contradiction:

Improveservice level agreement complianceVSAvoidenergy waste

Core Design Contradiction:

ReliabilityVSLoss of energy

Solution Approach 1:

The patent implements dynamic resource provisioning that automatically adjusts resource allocation based on real-time workload demands. The system transitions from static over-provisioning to dynamic adaptation, where resources are allocated proactively before demand occurs and scaled back when demand decreases, resolving the contradiction between maintaining SLA compliance and reducing energy waste.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system performs preliminary resource allocation based on predicted future workload demands rather than reacting to current demand. By using prediction algorithms to anticipate resource needs before they occur, the system can prepare resources in advance while avoiding the continuous over-provisioning that causes energy waste, thus improving both reliability and efficiency.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If static prediction models are used, then manufacturing precision of prediction is improved for stable conditions, but adaptability to behavioral changes deteriorates

Engineering Contradiction:

Improveprediction accuracyVSAvoidadaptability to behavioral changes

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent implements a feedback mechanism where the prediction model continuously learns from actual system behavior and adjusts its parameters accordingly. The system compares predicted resource demands with actual demands, uses this feedback to refine prediction accuracy, and adapts to behavioral changes over time, resolving the contradiction between maintaining precision for stable conditions and adapting to changes.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system dynamically changes prediction model parameters based on observed system behavior and workload patterns. By adjusting parameters such as prediction windows, confidence intervals, and model weights in response to changing conditions, the system maintains high prediction accuracy for stable conditions while becoming adaptable to behavioral changes.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If environment-specific or application-specific solutions are used, then prediction accuracy for particular cases is improved, but versatility across different systems deteriorates

Engineering Contradiction:

Improveprediction accuracyVSAvoidversatility

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent implements a universal prediction framework that can be applied across different virtualized systems, environments, and applications. The system uses generalized prediction algorithms and adaptive learning mechanisms that automatically adjust to specific system characteristics without requiring environment-specific customization, thus achieving both versatility and maintained prediction accuracy across diverse contexts.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Reliability

If proactive resource allocation is implemented, then reliability of meeting future demands is improved, but device complexity increases

Engineering Contradiction:

Improvedemand fulfillmentVSAvoidsystem complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements a self-service resource allocation system where the virtualized environment automatically monitors its own workload patterns, predicts future resource needs, and allocates resources without external intervention. The system uses self-learning algorithms that automatically adapt to changing conditions, reducing the need for complex manual configuration and management while maintaining high reliability in meeting future demands.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS12136040B2Resource needs prediction in virtualized systems: generic proactive and self-adaptive solution

Publication Date: 2024.11.05 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)

AI summary

A method for real-time prediction of resource consumption by a system is provided that includes determining a real-time prediction of resource demand by the system. A Genetic Algorithm (GA) is used to dynamically determine an optimal size of a sliding window and an optimal number of predicted data within the real-time prediction of the resource demand. The data within the real-time prediction of the resource demand is adjusted based on an estimated probability of prediction errors and a variable padding, which is based on a mean of at least one previous standard deviation of the predicted data within the real-time prediction of the resource demand.