Hypervisor Resource Fold Factor for VM Core Locality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for assigning physical resources to virtual machines in data processing systems often lead to performance degradation due to inefficient locality of processor cores and memory, as they lack precise anticipation of resource usage, resulting in suboptimal allocation and increased contention among virtual processors.
Innovation Solution
The introduction of resource fold factors, such as processor and memory fold factors, allows hypervisors to anticipate and optimize the allocation of physical resources by specifying a percentage of anticipated usage, thereby improving locality and reducing contention among virtual machines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional methods for assigning physical resources to virtual machines are used, then resource allocation is simple, but performance degrades due to inefficient locality of processor cores and memory
Solution Approach 1:
The patent applies preliminary action by calculating and storing fold factor values before virtual machine deployment. The system pre-computes the ratio of maximum to minimum resource usage for each virtual machine based on historical data, and stores these fold factor values in a database. This allows the hypervisor to make informed placement decisions about physical resource allocation without complex real-time calculations, thereby improving virtual machine performance while maintaining manageable system complexity.
2Productivity
If physical resources are allocated without fold factor optimization, then allocation process is fast, but resource locality between processor cores and memory is inefficient
Solution Approach 1:
The system performs preliminary calculations of fold factor values based on historical resource usage data before deployment. This pre-computation stores the anticipated resource usage patterns, allowing the hypervisor to quickly make placement decisions that optimize processor-core-to-memory locality without time-consuming real-time analysis, thus improving access efficiency while keeping allocation time minimal.
Solution Approach 2:
The patent enables the hypervisor to skip complex real-time resource usage analysis by utilizing pre-calculated fold factor values. The system rushes through the placement decision process by directly applying these stored fold factor values to determine optimal physical resource allocation, thereby achieving both fast allocation and optimized resource locality simultaneously.
3Productivity
If virtual machines are placed without considering fold factors, then placement is straightforward, but resource fragmentation increases
Solution Approach 1:
The system calculates fold factor values in advance that represent the ratio of maximum to minimum resource usage for each virtual machine. These pre-computed values are stored and used by the placement strategy to make informed decisions about physical resource allocation. This approach reduces resource fragmentation by anticipating usage patterns while keeping the placement strategy relatively simple and straightforward to implement.
Data Source
AI summary
A technique for assigning physical resources of a data processing system to a virtual machine (VM) includes reading, by a hypervisor executing on the data processing system, a fold factor attribute for the VM. The fold factor attribute defines an anticipated usage of physical resources of the data processing system by the VM. The technique also includes mapping based on a value of the fold factor attribute, by the hypervisor, allocated virtual processors of the VM to the physical resources to maximize processor core access to local memory for ones of the allocated virtual processors that are anticipated to be utilized.