Virtual Networking Device NUMA Node Migration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Virtual machines operating on host machines with Non-Uniform Memory Access (NUMA) architecture face performance issues when NUMA nodes associated with a guest operating system change, leading to inefficient data transfer and increased computing resources consumption due to the need for virtual NIC reconfiguration and potential updates in ARP tables.
Innovation Solution
The implementation of multiple virtual networking devices, each associated with a different NUMA node but sharing the same physical address, allows seamless migration of guest operating systems between hypervisors without updating ARP tables or instantiating new virtual devices, ensuring efficient data transfer and reduced resource consumption by using virtual devices corresponding to the new NUMA nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machines operate on host machines with NUMA architecture using traditional single virtual networking device per hypervisor, then network connectivity is maintained, but performance degrades when NUMA nodes change due to inefficient data transfer and increased resource consumption
Solution Approach 1:
The patent segments the virtual networking functionality by creating multiple virtual networking devices, where each device is associated with a specific NUMA node. This segmentation allows the guest operating system to select the appropriate virtual networking device based on its current NUMA node association, ensuring efficient data transfer and reducing computing resource consumption during NUMA node changes.
Solution Approach 2:
The patent implements dynamic virtual networking device selection where the guest operating system can dynamically choose which virtual networking device to use based on its current NUMA node affiliation. This dynamic approach ensures that network operations always use the optimal device associated with the current NUMA node, maintaining high performance during migrations between hypervisors or NUMA nodes.
2Reliability
If virtual networking devices are reconfigured during NUMA node changes, then network connectivity is maintained, but network reconfiguration overhead and ARP table updates are required
Solution Approach 1:
The patent performs preliminary action by pre-associating each virtual networking device with a specific NUMA node before migration occurs. Multiple virtual networking devices are prepared in advance, each tied to a different NUMA node. When the guest operating system migrates to a new NUMA node, it can immediately switch to the pre-associated virtual networking device without requiring reconfiguration or ARP table updates, thus maintaining network connectivity without time loss.
3Adaptability or versatility
If multiple virtual networking devices are created for different NUMA nodes, then seamless migration is enabled, but device complexity increases
Solution Approach 1:
The patent applies universality by creating multiple virtual networking devices that share the same physical address and can serve the same guest operating system at different times. Each virtual networking device is multi-functional, capable of handling network traffic for its associated NUMA node. This universal approach enables seamless migration while managing complexity through shared physical addressing and unified device functionality across different NUMA contexts.
Data Source
AI summary
A method includes, migrating a guest operating system from a first hypervisor to a second hypervisor, the first hypervisor utilizing a first virtual networking device associated with a first Non-Uniform Memory Access (NUMA) node, the first virtual networking device corresponding to a physical network device. The method further includes, after the migrating, with the second hypervisor, providing to the guest operating system, a second virtual networking device associated with a second NUMA node, the second virtual networking device having a same physical address as the first virtual networking device, the second virtual networking device corresponding to the physical network device.


