Speculative VM Resource Scheduling via Lockstep Duplication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual machine schedulers base migration decisions on past resource usage, leading to potential incorrect decisions and costly resource wastage and time delays due to the time-consuming nature of VM migration.
Innovation Solution
A speculative virtual machine resource scheduler creates duplicate VMs on secondary hosts and runs them in lockstep with the primary VM, allowing real-time resource monitoring to determine the most advantageous host for continued operation, thereby enabling instantaneous migration based on current resource availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If VM migration decisions are based on past resource usage, then scheduling simplicity is maintained, but decision accuracy deteriorates leading to incorrect migration decisions
Solution Approach 1:
The system performs preliminary actions by creating duplicate VM instances on secondary hosts before making final migration decisions. These duplicate VMs are run in lockstep with the primary VM, allowing the system to evaluate current resource conditions and predict future performance without committing to migration until the decision is confirmed advantageous.
Solution Approach 2:
The system creates duplicate copies of the VM on secondary hosts to evaluate migration potential. These copies allow the system to test resource compatibility and performance predictions without affecting the primary VM operation, enabling accurate decision-making based on actual current conditions rather than historical data alone.
2Speed
If VM migration is performed based on historical data, then processing speed is maintained, but resource allocation efficiency deteriorates due to time delays
Solution Approach 1:
The system performs preliminary resource evaluation and duplicate VM execution in parallel with the primary VM operation. This allows migration decisions to be made rapidly based on current real-time conditions rather than historical data, simultaneously improving processing speed and resource allocation efficiency.
Solution Approach 2:
The system continuously monitors resource conditions and performance metrics of duplicate VMs, using this feedback to make informed migration decisions. This real-time feedback mechanism enables both fast processing and efficient resource allocation by base(ing) decisions on current actual conditions rather than outdated historical information.
3Measurement precision
If duplicate VMs are created and run in lockstep for speculative scheduling, then migration decision accuracy is improved, but device complexity increases
Solution Approach 1:
The system creates duplicate copies of the VM to evaluate migration options. While this increases complexity, the copies are run in lockstep using efficient synchronization mechanisms and can be terminated once a decision is made, limiting the duration and impact of the additional complexity.
Solution Approach 2:
The duplicate VMs are run in lockstep with the primary VM, allowing continuous evaluation of resource conditions without interrupting service. This continuous parallel operation enables accurate migration decisions while maintaining system availability and minimizing the functional impact of the increased complexity.
4Productivity
If real-time resource monitoring is implemented for speculative scheduling, then resource allocation efficiency is improved, but energy consumption increases
Solution Approach 1:
The system implements partial monitoring by focusing resources on monitoring only those hosts and resources relevant to potential migration decisions, rather than monitoring the entire infrastructure continuously. This selective monitoring approach maintains resource allocation efficiency while reducing overall energy consumption compared to comprehensive continuous monitoring.
Solution Approach 2:
The system performs preliminary evaluation using duplicate VMs that consume resources locally on secondary hosts, reducing the need for continuous real-time monitoring and data transfer across the network. This preliminary action approach improves resource allocation efficiency while minimizing the energy required for ongoing monitoring.
Data Source
AI summary
A speculative virtual machine (“VM”) resource scheduler for a VM that is hosted by a primary host server initiates the creation of at least one duplicate VM on a secondary host server. The VM and the duplicate VM are run substantially in lockstep while the scheduler monitors resources on the primary host server and the secondary host server. The scheduler then determines an advantageous host server among the primary host server and the secondary host server and, based at least in part on the determination, initiates the powering down of either the VM on the primary host server or the duplicate VM on the secondary host server.