Virtual Machine Migration via Resource Sharing and Load Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualization systems, existing load balancing algorithms often fail to optimize resource utilization across all hosts effectively, leading to unbalanced loads and inefficient resource sharing due to exclusion criteria that prevent optimal placement of virtual machines.
Innovation Solution
A system-wide target host optimization mechanism that combines load balancing with resource sharing considerations, using a scheduling algorithm to determine the 'best overall placement' of virtual machines across candidate hosts based on current and historical resource utilization ratios, including CPU, memory, I/O, and network traffic, to maximize resource sharing and balance loads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If existing load balancing algorithms use exclusion criteria to select target hosts, then host selection is simplified, but resource utilization optimization is compromised
Solution Approach 1:
The patent changes the parameters used for host selection from simple exclusion criteria to a comprehensive scoring system that includes resource sharing level, expected load per resource, and historical utilization ratios. This allows the system to evaluate multiple dimensions of resource utilization simultaneously, resolving the contradiction between simple selection and optimized utilization.
Solution Approach 2:
The patent implements a feedback mechanism by considering historical resource utilization ratios and expected future load patterns when selecting target hosts. This feedback loop enables the system to learn from past performance and make more informed decisions, improving resource utilization while maintaining operational simplicity through automated scoring.
2Reliability
If VMs are migrated to balance load on hosts, then system health is improved, but service level disturbances occur during migration
Solution Approach 1:
The patent performs preliminary actions by calculating expected load per resource and determining optimal target hosts before actual VM migration occurs. This advance planning allows the system to prepare target hosts with sufficient resources and minimize disruption during the actual migration process, improving system health while reducing service disturbances.
Solution Approach 2:
The patent applies beforehand cushioning by selecting target hosts with adequate resource headroom and by considering historical utilization patterns that indicate stable performance. This cushioning approach ensures that migrations are planned to destinations that can absorb the incoming load without causing service level disturbances, thus protecting system health while minimizing harmful effects.
3Productivity
If resource sharing level is considered in host selection, then resource utilization is optimized, but selection complexity increases
Solution Approach 1:
The patent segments the complex selection process into distinct scoring components: resource sharing level assessment, expected load per resource calculation, and historical utilization ratio evaluation. By breaking down the complexity into manageable segments, the system can optimize resource utilization through structured analysis while keeping the overall selection process manageable and systematic.
Data Source
AI summary
A mechanism for system-wide target host optimization in load balancing virtualization systems is disclosed. A method of the invention includes detecting a condition triggering a load balancing operation, identifying a plurality of candidate target host machines to receive one or more operating virtual machines (VMs) to be migrated, determining a load per resource on each identified candidate target host machine, and scheduling all operating VMs among all of the identified candidate target host machines in view of an expected load per resource on each candidate target host.