VM Allocation Management for Data Center Power Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face high electricity costs due to inefficient use of physical server resources, with existing allocation techniques failing to adequately reduce electricity consumption despite efforts to reallocate resources based on usage rates.
Innovation Solution
A method involving an allocation management apparatus that determines idle virtual machines and migrates them to a separate server, while allocating active virtual machines to application servers, thereby reducing the number of powered-on servers and optimizing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machines are allocated based on usage rate alone, then resource allocation efficiency improves, but electricity consumption reduction is insufficient
Solution Approach 1:
The patent segments virtual machines into two distinct categories: active VMs (executing request-response processing) and idle VMs (not executing such processing). This segmentation enables differentiated allocation strategies where idle VMs can be concentrated on dedicated idle VM host apparatuses, allowing those host apparatuses to be powered down or reduced in power when idle VMs predominate, thereby directly reducing electricity consumption while maintaining resource allocation efficiency.
Solution Approach 2:
The patent extracts idle virtual machines from active VM host apparatuses and relocates them to dedicated idle VM host apparatuses. This extraction allows the active VM hosts to operate with higher efficiency by eliminating idle VM overhead, while the idle VM hosts can be independently managed for power optimization, thus addressing both resource allocation efficiency and electricity consumption reduction.
2Reliability
If more servers are kept powered on to maintain resource availability, then service reliability improves, but electricity consumption increases
Solution Approach 1:
By segmenting VM host apparatuses into active VM hosts and idle VM hosts, the patent enables independent power management. Idle VM hosts can be powered down or placed in low-power states when they are not needed, while active VM hosts maintain full power for service availability. This segmentation resolves the contradiction by allowing servers to be powered off without compromising service reliability.
Solution Approach 2:
The patent introduces a VM allocation management apparatus as an intermediary that intelligently manages the allocation and migration of virtual machines between active and idle host apparatuses. This intermediary ensures that service availability is maintained by properly orchestrating VM placement, while enabling power consumption reduction by consolidating idle VMs and allowing their host apparatuses to be powered down.
3Loss of energy
If virtual machines are consolidated to reduce server count, then electricity consumption decreases, but performance degradation occurs
Solution Approach 1:
The patent segments consolidation into two separate consolidation processes: active VMs are consolidated onto active VM host apparatuses that maintain full processing capability, while idle VMs are consolidated onto idle VM host apparatuses that can be powered down. This segmentation prevents performance degradation by ensuring that active processing VMs are not consolidated in a way that compromises their execution environment.
Solution Approach 2:
The patent applies different quality standards to different VM host apparatuses based on their function. Active VM host apparatuses maintain high-performance specifications suitable for executing request-response processing, while idle VM host apparatuses can have reduced specifications since they only need to maintain VM states without active processing. This local quality differentiation enables consolidation while preserving performance where needed.
Data Source
AI summary
An allocation control method includes: determining, for each of plural virtual machines executed on a first information processing apparatus or a second information processing apparatus, whether the virtual machine is executing a first processing that is performed in response to a request from an external apparatus; first allocating, to the first information processing apparatus, first virtual machines, which are executing the first processing, among the plural virtual machines; and second allocating, to the second information processing apparatus, second virtual machines, which are not executing the first processing, among the plural virtual machines.