Virtual Machine Scheduling via Rack Power Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual machine scheduling methods in cloud computing that rely solely on CPU parameters lead to unbalanced power consumption, resulting in inefficient resource utilization and increased costs due to uneven power distribution across data center racks.
Innovation Solution
A method and system for scheduling virtual machines that consider energy consumption data of racks and host computers, determining power consumption data of virtual machines, and dispatching them to target host computers based on real-time energy and power consumption metrics, thereby balancing power distribution and improving resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If virtual machine scheduling is implemented based on CPU parameters only, then scheduling simplicity is maintained, but power consumption becomes unbalanced across racks
Solution Approach 1:
The patent changes the scheduling parameters from CPU-only metrics to a comprehensive set including power consumption, energy efficiency ratios, and rack-level power metrics. This allows the system to consider multiple dimensions (power consumption, resource utilization) when selecting target host computers, thereby resolving the contradiction between scheduling simplicity and power consumption balance.
2Productivity
If more servers are racked to fully share costs, then resource utilization improves, but power consumption density increases and causes scheduling waste
Solution Approach 1:
The patent implements dynamic scheduling that adapts to changing power consumption conditions. The system continuously monitors power metrics and adjusts virtual machine placement decisions in real-time, allowing the data center to maximize resource utilization while dynamically responding to power consumption constraints and avoiding scheduling waste.
Solution Approach 2:
The patent applies different scheduling strategies to different racks based on their local power consumption characteristics. By evaluating rack-specific power metrics and energy efficiency ratios, the system optimizes virtual machine placement at the rack level, allowing high-utilization racks to accept more workloads while directing new virtual machines to racks with better power headroom.
3Reliability
If rack density is adjusted based on experience values with reserved space, then system stability is maintained, but cost efficiency decreases due to underutilization
Solution Approach 1:
The patent implements feedback mechanisms that continuously monitor power consumption, resource utilization, and system performance. This real-time feedback allows the system to dynamically adjust rack density and virtual machine placement decisions, replacing static experience-based reservations with data-driven optimization that maintains system stability while improving cost efficiency.
Data Source
AI summary
A method and a system for scheduling a virtual machine are disclosed. The method includes: obtaining energy consumption data of a rack and energy consumption data of host computers, wherein the rack is used to place the host computers; determining power consumption data of a virtual machine; determining a target host computer corresponding to the virtual machine according to the energy consumption data of the rack, the energy consumption data of the host computers, and the power consumption data of the virtual machine; and dispatching the virtual machine to the target host computer. The present disclosure solves the technical problems of unbalanced power consumption scheduling caused by virtual machine scheduling methods that are implemented only according to CPU parameters in the existing technologies.