Predictive Virtual Machine Migration and Host Power Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional Distributed Resource Scheduling (DRS) and Distributed Power Management (DPM) operate in a reactive mode, leading to negative performance impacts when launching VM migrations and host power-ons or power-offs in response to increasing VM demand.
Innovation Solution
A predictive approach that analyzes current workload to forecast future demands, recommending changes to the virtual computing environment and performing a cost-benefit analysis to determine if these changes will improve performance, with the VM management center deciding whether to implement suggested changes based on their impact on current and future workloads.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reactive mode is used for DRS and DPM to respond to VM demand changes, then recommendations are justified by observed data, but launching VM migrations and host power-ons/offs during increasing VM demand causes negative performance impact
Solution Approach 1:
The patent applies preliminary action by performing VM migrations and host power operations before VM demand increases, rather than reacting after demand changes are observed. The predictive workload analysis identifies upcoming demand patterns, allowing the system to proactively relocate VMs to hosts that will have sufficient capacity when demand peaks, thereby avoiding performance degradation while maintaining recommendation validity
Solution Approach 2:
The patent implements dynamics by transitioning from a static reactive response model to a dynamic predictive model that continuously analyzes workload patterns and adapts migration timing. The system dynamically adjusts when to execute migrations based on predicted workload trajectories, enabling optimal performance by operating ahead of demand fluctuations rather than reacting to them
2Adaptability or versatility
If VM migrations are launched reactively based on current demand data, then resource allocation responds to actual conditions, but resource contention and latency increase during high-demand periods
Solution Approach 1:
The system performs VM migrations in advance of predicted demand increases, eliminating resource contention and latency that would occur during high-demand periods. By analyzing workload patterns and predicting future demand, the system proactively relocates VMs to ensure adequate resource availability before demand peaks, rather than attempting migrations during contention-prone periods
Solution Approach 2:
The patent applies skipping by executing VM migrations during low-demand periods when resource contention is minimal, effectively skipping over the high-demand periods when migrations would cause latency. The predictive model identifies optimal migration windows that avoid resource contention, allowing migrations to complete quickly without impacting VM performance
3Use of energy by stationary object
If host power-ons and power-offs are performed reactively, then energy management responds to current workload, but performance impact occurs when operations coincide with increasing VM demand
Solution Approach 1:
The patent applies preliminary action to host power management by predicting future workload patterns and scheduling host power-ons before VM demand increases require additional capacity. This ensures hosts are powered on and ready to receive migrated VMs without causing performance degradation, while power-offs are scheduled during predicted low-demand periods to minimize impact
Solution Approach 2:
The system uses feedback from predictive workload analysis to dynamically adjust host power management decisions. By continuously monitoring workload patterns and predicting future demand, the system receives feedback that informs optimal timing for power operations, balancing energy efficiency with performance requirements
Data Source
AI summary
A technique for managing distributed computing resources in a virtual computing environment is disclosed. In an embodiment, a method includes receiving a recommended change to a virtual architecture of a virtual computing environment; determining an impact on current workload in the virtual computing environment if the recommended change is performed; determining an impact on future workload in the virtual computing environment if the recommended change is performed; calculating a combined impact on current and future workload; determining if the combined impact is above or below a threshold; if the combined impact on current and future workload is below the threshold, do not perform the recommended change; and if the combined impact on current and future workload is above the threshold, perform the recommended change.