VM Placement Algorithm Minimizing Migration Frequency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face inefficiencies and increased operational complexities due to frequent virtual machine migrations, which can lead to data loss and resource wastage, particularly in high-density environments like hyperscale and mega data centers.
Innovation Solution
A method and apparatus for placing virtual machines on physical machines that calculates workload information, initial predicted virtual load, and search placement to minimize migrations, using objective functions that optimize workload distribution and reduce the number of virtual machine movements, thereby stabilizing machine operations and reducing power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machine migration is performed to balance physical machine loads, then load distribution is improved, but the number of migrations increases and data loss risk increases
Solution Approach 1:
The patent calculates predicted virtual load information in advance for a predetermined period, allowing the system to plan migrations before actual overload occurs. This preliminary prediction and planning action enables proactive load balancing while minimizing unnecessary migrations that would increase data loss risk.
Solution Approach 2:
The patent uses objective functions that incorporate predicted load information and migration history to guide placement decisions. The system continuously monitors actual workload and compares it with predictions, adjusting migrations based on this feedback loop to achieve better load distribution with fewer migrations.
2Productivity
If frequent migrations are performed to maintain balance, then load distribution is improved, but operational complexity and resource consumption increase
Solution Approach 1:
By calculating predicted virtual load information for a predetermined period in advance, the system can plan migrations proactively rather than reactively. This reduces the frequency of migrations needed to maintain balance, thereby lowering operational complexity and resource consumption.
Solution Approach 2:
The patent dynamically adjusts migration strategies based on predicted load patterns and actual workload conditions. The system adapts its migration behavior to match actual usage patterns, reducing unnecessary migrations and simplifying operational complexity while maintaining effective load distribution.
3Productivity
If migrations are performed many times to correct overload, then load balance is improved, but the number of migrations increases and resource waste increases
Solution Approach 1:
The system performs preliminary calculations of predicted virtual load information before making placement decisions. This allows the system to anticipate future load conditions and plan migrations that will achieve load balance with fewer iterations, reducing resource waste associated with repeated migrations.
Solution Approach 2:
The patent incorporates feedback from actual workload measurements and migration history into the placement algorithm. This feedback mechanism allows the system to learn from previous migrations and adjust future decisions to achieve load balance more efficiently, reducing the total number of migrations and associated resource waste.
Data Source
AI summary
A method of placing a virtual machine may include: calculating virtual machine workload information, which is a load of the physical machine applied by the virtual machine through a predetermined method of calculating a workload on the basis of information stored in storage and related to operations of the first physical machine and the second physical machine; calculating initial predicted virtual load information related to an initial predicted virtual load related to a load of the physical machine predicted due to the virtual machine for a predetermined period on the basis of the virtual machine workload; calculating search placement information related to a search placement which is a placement of the virtual machine on the first physical machine and the second physical machine through a predetermined method of calculating a search placement; and placing the virtual machine on the physical machine again on the basis of the selected search placement.