VM Assignment to Compute or Hyper-Converged Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional virtual computing systems face performance bottlenecks and network congestion due to the configuration of nodes providing either compute resources only or both compute and storage resources, leading to inefficient VM migration and increased network traffic.
Innovation Solution
Implementing a method to identify and rank virtual machines based on input-output patterns, assigning them to either compute-only (CO) or hyper-converged (HC) nodes to optimize performance, and monitoring changes in behavior to improve network efficiency and reduce load by ensuring data locality and minimizing network usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If VMs are migrated to minimize inter-VM communications over network, then network traffic reduction is expected, but network usage due to migration process increases
Solution Approach 1:
The system performs preliminary analysis of VM I/O patterns and communication behaviors before migration decisions are made. By pre-identifying VMs with high storage I/O requirements and their associated vDisks, the system prepares optimal migration targets in advance, reducing the actual migration workload and network usage during the migration execution phase.
Solution Approach 2:
The patent introduces an intermediary mechanism that analyzes I/O patterns and acts as a mediator between VM migration decisions and network resource allocation. This intermediary identifies VMs that would benefit from co-location with their vDisks and facilitates migrations that reduce long-term network traffic while managing migration overhead.
2Speed
If physical network card is saturated by co-locating VMs that send or receive considerable amount of data, then data locality is improved, but network congestion occurs
Solution Approach 1:
The system applies local quality by differentiating between different types of VMs based on their I/O patterns. VMs with high storage I/O requirements are co-located with their vDisks on the same node to achieve data locality and fast access, while VMs with lower storage needs can be placed on compute-only nodes. This selective co-location optimizes data access speed for relevant VMs without causing universal network congestion.
Solution Approach 2:
The patent changes the placement parameter (node type: CO vs HC) based on VM-specific I/O characteristics. By analyzing individual VM I/O patterns and dynamically adjusting placement decisions, the system achieves data locality for VMs that require it while avoiding network saturation that would result from universal co-location policies.
3Productivity
If VMs are migrated frequently to optimize performance, then system performance is improved, but network usage and migration overhead increase
Solution Approach 1:
The system performs preliminary analysis of VM I/O patterns and communication behaviors before migration decisions are made. By pre-identifying VMs with high storage I/O requirements and their associated vDisks, the system prepares optimal migration targets in advance, reducing the actual migration workload and network usage during the migration execution phase.
Solution Approach 2:
The patent implements a feedback mechanism that continuously monitors VM I/O patterns, network traffic, and system performance metrics. This feedback information is used to dynamically adjust migration decisions, ensuring that migrations are performed only when they will genuinely improve performance and not when they would merely increase migration overhead and network usage.
Data Source
AI summary
A system and method include assigning virtual machines (VMs) to any one of compute only (CO) and hyper converged (HC) nodes. The method includes identifying, by a management processor of a virtual computing system, input-output patterns for a plurality of virtual disks in the virtual computing system; ranking, by the management processor, the plurality of virtual machines according to the identified input-output patterns; assigning, by the management processor, the plurality of virtual machines to any one of CO and HC nodes that optimize performance of the virtual computing system; monitoring, by the management processor, performance of the plurality of virtual machines; and determining, by the management processor, a change in behavior of one of the plurality of virtual machines.