Hypervisor Partial Surprise Removal for VM Migration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtual machine migration, the removal of devices like PCI devices from a source host system to a destination host system is often delayed due to the need for explicit acknowledgment from the guest OS, which can be time-consuming, especially if the guest OS is slow or non-responsive, leading to potential data loss and network communication disruptions.
Innovation Solution
Implementing a partial surprise removal method where the hypervisor disables host system access to the device before migration, allowing data transfer without waiting for guest OS acknowledgment, ensuring data integrity and reducing migration delays by routing all requests through the hypervisor.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If explicit acknowledgment from guest OS is required before device removal, then data integrity is maintained, but migration time increases significantly
Solution Approach 1:
The hypervisor performs preliminary actions by disabling host system access to the device and routing requests through the hypervisor before migration begins. This allows the guest OS to continue operating normally while the hypervisor prepares for device removal, eliminating the need to wait for guest OS acknowledgment and significantly reducing migration time while maintaining data integrity through controlled access.
Solution Approach 2:
The hypervisor acts as an intermediary between the host system and the device during migration. By routing all requests through the hypervisor, it mediates access to the device, allowing the guest OS to operate independently while the hypervisor controls host system access. This intermediary role enables device removal without guest OS acknowledgment while preventing data loss.
2Reliability
If device removal is delayed waiting for guest OS acknowledgment, then data loss is prevented, but network communication may be disrupted
Solution Approach 1:
The hypervisor performs preliminary configuration by disabling host system access to the device and setting up request routing through the hypervisor before migration begins. This preliminary setup allows network communication to continue uninterrupted through the hypervisor's mediation, while the device can be removed without waiting for guest OS acknowledgment, preventing both data loss and communication disruptions.
Solution Approach 2:
The hypervisor serves as an intermediary that maintains network communication continuity during device removal. By routing all device requests through itself, the hypervisor ensures that network communication continues uninterrupted even as the device is being removed from the host system, eliminating the trade-off between data loss prevention and communication continuity.
3Productivity
If host system access to device is allowed during migration, then device operation continues normally, but data loss risk increases
Solution Approach 1:
The system segments access control by separating guest OS access from host system access. The guest OS continues to access the device normally through its driver, maintaining device operation continuity, while the hypervisor disables host system access and routes requests through itself. This segmentation allows continuous device operation while preventing data loss by controlling host system access.
Solution Approach 2:
The hypervisor acts as an intermediary that selectively controls access to the device. It allows guest OS requests to pass through while blocking host system access, thereby maintaining device operation continuity for the guest while preventing data loss from uncontrolled host access during migration.
Data Source
AI summary
A method includes disabling, by a hypervisor executed by a processing device of a source host system, host system access of a device associated with a virtual machine, wherein disabling the host system access prevents the virtual machine from directly accessing guest memory via the device and prevents the device from causing an interrupt. The method further includes migrating the guest memory associated with the virtual machine from the source host system to a destination host system. Upon migrating the guest memory associated with the virtual machine from the source host system to the destination host system, initiation of the virtual machine is caused on the destination host system.