Fuzzy Resource Allocation for Cloud Software Applications
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing resource allocation systems in cloud environments face challenges in automatically and adaptively managing resources for software applications with changing demands and competing priorities, often leading to inefficient resource utilization and potential overload.
Innovation Solution
A fuzzy-based control system that measures current resource usage, determines resource utilization values, and adjusts resource allocation by calculating scaling values based on relevance and criticality, ensuring that resource usage does not exceed adjustable thresholds, thereby optimizing resource allocation across software applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If self-organizing controllers with fixed performance measures are used to automatically adjust resource allocation, then the system can adapt to changing demands, but the rule base oscillates around ideal values with frequent modifications and little optimization potential
Solution Approach 1:
The patent applies dynamics by making the performance measure itself adaptive rather than fixed. The performance measure dynamically adjusts its target values based on current system state and learned patterns, allowing the controller to adapt to changing demands without oscillating around a static ideal value. This resolves the contradiction by enabling adaptability while maintaining stability through dynamic rather than rigid performance criteria.
Solution Approach 2:
The system employs self-service through machine learning components that automatically learn optimal performance measures and target values from historical data and current system behavior. Rather than requiring manual rule adjustment or oscillating around fixed targets, the system serves itself by continuously improving its performance criteria based on observed patterns, achieving both adaptability and stability.
2Adaptability or versatility
If multiple fuzzy-based devices with different target values are used to control resource allocation, then adaptability to different application requirements is improved, but system complexity increases
Solution Approach 1:
The patent applies universality by creating a multi-functional resource allocation system where a single integrated controller can handle multiple software applications with different requirements. The system uses a unified fuzzy logic framework that can dynamically adapt its performance measures and target values for different applications, eliminating the need for separate specialized controllers for each application while maintaining adaptability to diverse requirements.
Solution Approach 2:
The system resolves the complexity issue by dynamically changing parameters (performance measures and target values) rather than creating separate devices for different scenarios. The fuzzy-based controller adjusts its internal parameters based on the specific application being served, allowing one device to perform multiple functions with different characteristics without increasing structural complexity.
3Productivity
If resource allocation is continuously adjusted to meet changing application demands, then resource utilization efficiency is improved, but the risk of exceeding system capacity and causing overload increases
Solution Approach 1:
The patent applies feedback by continuously monitoring actual resource usage against dynamically determined target values and adjusting allocations accordingly. The system uses performance measures that incorporate feedback from system state and historical data to prevent over-allocation. This feedback mechanism ensures high resource utilization while maintaining reliability by adjusting allocations before capacity limits are exceeded.
Solution Approach 2:
The system uses preliminary action by proactively adjusting resource allocations based on predicted future demands and current trends rather than reacting only to immediate conditions. The machine learning components analyze patterns to anticipate resource needs and adjust allocations in advance, preventing both under-utilization and overload by taking preventive measures before capacity issues arise.
Data Source
Figure 1
Figure 2
Figure 3a~3b
AI summary
The invention relates to a method and system for fuzzy-based control of the allocation of resources to one or more software applications forming a group, wherein the current resource utilization by each software application of the group is measured for each resource, an instantaneous resource utilization value is determined for each software application depending on its current resource utilization of each resource, at least the software application with the highest resource utilization value is selected, and a scaling value is determined for at least one selected software application depending on its resource utilization value, a relevance value, a criticality value, and a system capacity for the group, with the criterion that the utilization of all resources by all software applications of the group does not exceed an adjustable threshold.