VM Disk Write Queueing for Application-Consistent Backup Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional virtual machine services face performance issues during snapshot operations, leading to sustained performance drops, and existing solutions for high availability, such as Remus, involve complex software reengineering and hardware deployment.
Innovation Solution
A system and method for marking application-consistent points-in-time by intercepting and replicating write requests for a virtual machine disk, queuing these requests to maintain an ordered sequence, and asynchronously backing up the disk to avoid prioritizing backup operations over others, thus reducing performance drops.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional snapshot operations are performed on virtual machine disks, then data backup is achieved, but performance drops significantly and lingers even after snapshot release
Solution Approach 1:
The system performs preliminary actions by queuing write requests before they reach the virtual machine disk. The I/O filter intercepts and queues write requests in an I/O filter queue, and the I/O daemon queues them again in an I/O daemon queue before processing. This preliminary queving ensures that when backup operations occur, the data is already organized and ready, preventing performance degradation during actual backup execution.
Solution Approach 2:
The patent introduces intermediary components (I/O filter and I/O daemon) between the guest virtual machine and the virtual machine disk. These intermediaries manage write requests asynchronously, allowing the backup process to operate independently without directly impacting the performance of the virtual machine. The intermediary layer buffers and manages data flow, isolating performance impacts.
2Speed
If backup operations are prioritized over other operations, then backup speed increases, but performance of other virtual machine operations deteriorates
Solution Approach 1:
The system maintains continuous useful action by allowing write requests to be queued and processed asynchronously without interruption to the virtual machine operations. The I/O filter and I/O daemon continuously manage the queue of write requests, ensuring that backup operations proceed at optimal speed while the virtual machine continues its normal operations without performance degradation. Both processes run concurrently without one blocking the other.
Solution Approach 2:
The patent implements dynamic processing by allowing the backup system to adaptively handle write requests based on current system conditions. The asynchronous queving mechanism dynamically adjusts the processing order and timing of write requests, enabling the system to optimize backup speed when resources are available while maintaining virtual machine performance when resources are constrained.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosed computer-implemented method for marking application-consistent points-in-time may include intercepting, by an I/O filter, a write request from a guest virtual machine to a virtual machine disk and queueing the write request in an I/O filter queue. The method may include sending the write request to the virtual machine disk and receiving a write completion message from the virtual machine disk. The method may also include sending, in response to the write completion message, the write request to an I/O daemon, and queueing the write request in an I/O daemon queue. The method may further include sending the write completion message to the guest virtual machine, and sending the write request to a backup gateway such that the backup gateway mimics writes to the virtual machine disk. Various other methods, systems, and computer-readable media are also disclosed.