Virtual Machine Allocation in NUMA Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In systems employing Non-Uniform Memory Access (NUMA), it is challenging to evaluate the influence of node errors on multiple virtual machines and optimize the allocation of virtual machines to minimize the impact of such errors, due to the complexity of evaluating availability and process performance across multiple nodes.
Innovation Solution
An information processing apparatus calculates an availability evaluation value indicating the degree of deviation in virtual machine distribution across nodes, allowing for the evaluation of error influence and optimal virtual machine allocation by specifying the number of virtual machines on each node and determining the best combination patterns for cores and memories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machines are allocated across multiple nodes in a NUMA system, then system productivity and resource utilization are improved, but the complexity of evaluating and managing error influence across nodes increases
Solution Approach 1:
The patent replaces complex manual evaluation mechanisms with an automated information processing apparatus that calculates availability evaluation values using standardized formulas. This substitution transforms the mechanical process of error influence assessment into an automated computational system, reducing evaluation complexity while maintaining productivity benefits across NUMA nodes
Solution Approach 2:
The patent implements feedback mechanisms where availability evaluation values are continuously calculated and used to adjust virtual machine allocation decisions. The system feeds back error influence assessments to the allocation process, enabling dynamic optimization that maintains high productivity while managing complexity through iterative improvement
2Reliability
If virtual machines are concentrated on fewer nodes, then error influence evaluation becomes simpler, but system reliability decreases due to reduced fault tolerance
Solution Approach 1:
The patent changes the parameter of availability evaluation from a qualitative assessment to a quantitative metric calculated using standardized formulas. By transforming error influence evaluation into measurable parameters (availability evaluation values), the system can objectively compare different allocation scenarios, improving reliability through data-driven decisions while managing complexity via standardized calculations
Solution Approach 2:
The patent performs preliminary calculation of availability evaluation values before finalizing virtual machine allocation decisions. By pre-assessing error influence on potential allocation configurations, the system identifies optimal allocations that maximize reliability while minimizing management complexity, avoiding the need for complex post-allocation adjustments
3Productivity
If detailed error influence evaluation is performed across all nodes, then allocation optimization improves, but the time and computational resources required increase
Solution Approach 1:
The patent applies partial evaluation by focusing computational resources on calculating availability evaluation values for critical nodes and allocation scenarios rather than performing exhaustive analysis of all possible configurations. This selective approach achieves sufficient allocation optimization while reducing evaluation time and computational overhead
Solution Approach 2:
The patent segments the error influence evaluation process into modular components that can be calculated independently for each node and then combined. By dividing the complex evaluation into smaller, manageable segments (node-level availability calculations), the system achieves comprehensive optimization without requiring prohibitive computational time
Data Source
AI summary
An information processing apparatus includes a first memory, and a processor coupled to the first memory and configured to: specify a number of virtual machines executed on each node of a plurality of nodes on an information processing system that performs as a plurality of virtual machines, and calculate a value indicating a degree of deviation of the number of the virtual machines between the plurality of nodes.