Virtual Disk SCSI Reservation Notification via File System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing virtual machine architectures fail to fully virtualize SCSI-2 reservations, leading to inconsistent and inaccurate locking of shared resources in clustering applications, which are critical for high availability and failover scenarios.
Innovation Solution
Implementing a method to fully virtualize SCSI-2 reservations by initializing virtual disks on a physical storage medium, managing reservation status through metadata, and using SCSI-3 commands to notify and update reservation status across distributed applications, ensuring that only authorized virtual machines access shared resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If virtual machine architecture is used to consolidate passive nodes, then hardware utilization improves, but SCSI-2 reservation virtualization becomes inconsistent and inaccurate
Solution Approach 1:
The patent introduces a file system as an intermediary layer between virtual machines and physical storage. The file system implements a notification mechanism using file metadata and registration keys that mediates SCSI-2 reservation semantics, translating them into file system operations that work correctly in virtualized environments. This intermediary resolves the contradiction by providing accurate reservation tracking without requiring changes to the underlying virtual machine architecture.
Solution Approach 2:
The patent changes the parameter representation of reservations from SCSI-2 protocol-specific fields to file system metadata attributes. By storing reservation information as file attributes and using file system commands rather than SCSI commands, the system transforms the reservation mechanism into one that is inherently compatible with virtualized storage, maintaining both hardware consolidation benefits and reservation accuracy.
2Adaptability or versatility
If SCSI-2 reservations are virtualized in existing VM architectures, then resource locking capability improves, but notification accuracy to other virtual machines deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the file system monitors registration keys and sends notifications to virtual machines when reservation status changes. When a VM registers for notifications on a file, the file system tracks this registration and actively notifies the VM when another VM changes reservations on that file, ensuring accurate and timely notification in virtualized environments.
Solution Approach 2:
The file system acts as an intermediary that captures and translates SCSI-2 reservation events into file system notification events. This intermediary layer ensures that reservation changes are accurately tracked and communicated to all interested virtual machines, resolving the notification accuracy problem while maintaining full resource locking capability.
3Ease of operation
If cluster software is designed for real machines, then SCSI-2 command handling is straightforward, but virtual machine resource isolation becomes insufficient
Solution Approach 1:
The patent creates a copy of the reservation notification mechanism at the file system layer, mirroring the SCSI-2 reservation semantics in the virtualized environment. By copying the essential reservation tracking and notification functionality into the file system, the patent maintains SCSI command handling simplicity for cluster software while adding the necessary virtualization layer for proper resource isolation.
Data Source
AI summary
A method for implementing virtual disk reservations on a storage medium for multiple distributed applications. The method includes initializing a plurality of virtual disks, wherein each of the virtual disks is instantiated on a physical storage medium. A reservation status is determined for each of the virtual disks and a first distributed application that uses at least one virtual disk is executed. Upon an update of the reservation status for the at least one virtual disk by the first distributed application, a notification indicator is set to notify a second distributed application of the update. The update is subsequently recognized and the notification indicator is cleared in order to continue I/O for the first distributed application and the second distributed application.