Adaptive Resource Dimensioning for Cloud Virtualization Bins
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for resource dimensioning in multi-layer cloud stacks face challenges in efficiently assigning workloads due to recurring bin packing problems, where items of different sizes need to be packed into flexible bins, leading to overprovisioning or underprovisioning of resources, and existing algorithms fail to consider dynamic VM capacities and affinity rules.
Innovation Solution
A method for adaptive resource dimensioning that creates a dynamic model for workload assignment, allowing for the creation of new bins with dimensioned resource capacities to accommodate workloads, and adjusts resource requirements dynamically, considering hierarchical relationships between virtualization layers, thereby optimizing resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If fixed-size bins are used for workload assignment, then resource allocation is simple, but resource utilization is inefficient leading to overprovisioning or underprovisioning
Solution Approach 1:
The patent applies dynamics by transitioning from fixed-size bins to flexible bins whose capacities can be dynamically adjusted. The bin packing algorithm dynamically determines optimal bin sizes and capacities based on workload requirements, enabling efficient resource utilization while maintaining allocation simplicity through automated decision-making.
2Productivity
If dynamic bin capacities are introduced to improve resource utilization, then resource allocation complexity increases
Solution Approach 1:
The patent implements self-service through an automated bin packing algorithm that independently determines optimal bin capacities and assignments without manual intervention. The system self-adjusts bin sizes based on workload characteristics, eliminating the need for complex manual configuration while achieving efficient resource utilization.
Solution Approach 2:
The patent applies parameter changes by dynamically modifying bin capacity parameters based on workload requirements. The algorithm adjusts bin sizes, capacities, and configurations as parameters to optimize resource allocation, transforming the static bin parameter into a dynamic variable that adapts to changing conditions.
3Ease of manufacture
If existing bin packing algorithms are used, then implementation is straightforward, but they fail to consider dynamic VM capacities and affinity rules
Solution Approach 1:
The patent applies universality by designing a bin packing algorithm that can handle multiple types of constraints and requirements simultaneously. The algorithm is versatile enough to accommodate dynamic VM capacities, affinity rules, and various workload characteristics in a unified framework, making it applicable to diverse cloud infrastructure scenarios.
Data Source
AI summary
A method is provided performed by a network node for adaptive resource dimensioning for a cloud stack having a plurality of virtualization layers with resource dependencies between the plurality of virtualization layers in an infrastructure. The method includes assigning a workload to a bin in the infrastructure based on one of (i) the bin has a resource capacity that supports a dimension of the workload, and (ii) when the resource capacity of the bin is insufficient (a) create a new bin, and (b) assign the workload to the new bin having a dimensioned resource capacity. The method further includes outputting information representing one of the assigned bin and the assigned new bin for the workload, and a corresponding resource capacity of the assigned bin or the dimensioned resource capacity of the new bin.