Predictive Cloud Workload Migration for Cost and Carbon Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Cloud architectures are static and inefficient due to reliance on resources at design time, lacking flexibility and sustainability, and result in high costs and carbon footprints.

Innovation Solution

A system and method using mathematical models to estimate resource requirements and availability, enabling automatic workload migration to cost-effective and sustainable resources like spot instances and serverless instances based on historical usage patterns and carbon footprint analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Stability of the object's composition

If cloud architectures use static resources at design time, then architecture stability is maintained, but resource efficiency and cost-effectiveness deteriorate

Engineering Contradiction:
Improvearchitecture stabilityVSAvoidresource efficiency
Core Design Contradiction:
Stability of the object's compositionVSProductivity

Solution Approach 1:

The patent implements dynamic cloud architecture that automatically adjusts resource allocation based on real-time workload conditions. The system transitions from static design-time resource configuration to runtime dynamic provisioning, enabling the cloud infrastructure to adapt resource levels to actual demand patterns and optimize cost-efficiency while maintaining service quality.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system dynamically changes key parameters including resource allocation levels, provisioning rates, and capacity planning variables based on monitored workload metrics. This parameter adaptation enables the architecture to respond to changing conditions without requiring complete redesign, resolving the contradiction between stability and efficiency.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If cloud resources are over-provisioned to ensure availability, then service reliability is improved, but cost and carbon footprint increase

Engineering Contradiction:
Improveservice availabilityVSAvoidcarbon footprint
Core Design Contradiction:
ReliabilityVSLoss of energy

Solution Approach 1:

The patent implements feedback mechanisms that continuously monitor workload demand, resource utilization, and availability metrics. This feedback loop enables the system to dynamically adjust resource provisioning to match actual demand, avoiding both over-provisioning (which increases carbon footprint) and under-provisioning (which compromises availability), thus resolving the contradiction between reliability and energy efficiency.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system employs self-service automation where the cloud infrastructure autonomously manages resource allocation based on monitored conditions. This eliminates the need for manual over-provisioning while maintaining service availability through automated scaling and load balancing, reducing unnecessary energy consumption and carbon emissions.

Inventive Principle:
Principle #25Self-service

3Productivity

If cloud infrastructure is optimized for cost-efficiency, then operational expenses are reduced, but service quality and performance may deteriorate

Engineering Contradiction:
Improvecost-efficiencyVSAvoidservice quality
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system dynamically adjusts operational parameters such as resource allocation levels, instance types, and provisioning strategies based on real-time service quality metrics and workload characteristics. This enables cost-optimized resource selection that maintains required service quality standards, resolving the contradiction between cost-efficiency and service quality.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4610824A1System and method for managing computing resources
Publication Date: 2025.09.03 SIEMENS AG
  • EP4610824A1 patent drawingFigure 1A
  • EP4610824A1 patent drawingFigure 1B
  • EP4610824A1 patent drawingFigure 2

AI summary

A method (300) for managing computing resources (122) for workload is disclosed. The method comprises estimating (302), by one or more processors (132) using a first mathematical model, a time period indicative of a minimum requirement of the computing resources for the workload. Further, the method comprises determining (304), by the one or more processors (132) using a second mathematical model, a probability of availability of a first type of computing resources (122a) during the estimated time period of the minimum requirement. Further, the method comprises comparing (306) the determined probability of availability of the first type of computing resources (122a) with an availability threshold. Furthermore, the method comprises migrating (308) the workload to the first type of computing resources (122a) at the estimated time period upon determining that the determined probability of availability is meeting the availability threshold.