Virtual Disk Snapshot Conversion Between COW and Redo
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional storage systems employing LUNs or FS volumes face scalability issues, resource contention, and inability to provide per-VM storage policies and service level agreements, as well as limitations in snapshotting and data services, particularly with virtualized environments where multiple VMs share storage resources.
Innovation Solution
The solution involves converting between copy-on-write (COW) and redo-based technologies by 'swizzling' disk descriptor files to enable snapshotting and reverting virtual disks, allowing for efficient management of virtual volumes and snapshots across storage systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple VMs share storage resources through LUNs or FS volumes, then storage scalability is improved, but resource contention and inability to provide per-VM storage policies occur
Solution Approach 1:
The invention segments storage resources at the virtual disk level rather than at the LUN or volume level. Each virtual disk can be independently managed and allocated to specific VMs, enabling fine-grained resource control and eliminating contention at the coarser LUN/volume level while maintaining scalability.
Solution Approach 2:
The invention implements per-VM storage policies and service level agreements by assigning local qualities to each virtual disk. Different VMs can have different storage performance characteristics, policies, and resource allocations tailored to their specific needs, rather than sharing uniform LUN-level resources.
2Ease of operation
If conventional snapshotting is used at LUN or volume level, then storage management is simplified, but per-VM snapshot capability and data services are limited
Solution Approach 1:
The invention segments snapshot capability at the virtual disk level, allowing individual snapshots to be taken for each virtual disk independently. This enables per-VM snapshotting and selective data services without requiring complex LUN or volume-level snapshot management.
3Productivity
If copy-on-write technology is used for snapshots, then snapshot efficiency is improved, but conversion to redo-based technology and seamless migration is difficult
Solution Approach 1:
The invention introduces a virtual disk layer as an intermediary between the storage system and VMs. This virtual disk can be presented in different snapshot technologies (COW or redo-based) depending on the underlying storage system capabilities, enabling seamless migration and conversion without affecting the VMs themselves.
Solution Approach 2:
The invention implements dynamic technology conversion where virtual disks can switch between COW and redo-based snapshot technologies based on storage system capabilities or migration requirements. This dynamic adaptation enables flexible migration paths and technology independence.
4Speed
If VMs are directly connected to physical disks, then storage access performance is improved, but scalability to 10,000+ VMs becomes unmanageable
Solution Approach 1:
The invention introduces a universal virtual disk layer that serves multiple functions: it provides block-level access for performance-critical workloads, enables virtualization for scalability, supports snapshotting, and facilitates resource allocation. This multi-functional layer resolves the contradiction between direct access performance and scalable management.
Data Source
AI summary
A framework for converting between copy-on-write (COW) and redo-based technologies is disclosed. To take a virtual disk snapshot, disk descriptor files, which include metadata information about data stored in virtual volumes (vvols), are “swizzled” such that the descriptor file for a latest redo log, to which IOs are currently performed, points to the base vvol of a COW-based vvol hierarchy. A disk descriptor file previously associated with the base vvol may also be updated to point to the vvol newly created by the snapshot operation. To revert to an earlier disk state, a snapshot may be taken before copying contents of a snapshot vvol of the COW-based vvol hierarchy to a base vvol of the hierarchy, thereby ensuring that the reversion can be rolled back if it is unsuccessful. Reference counting is performed to ensure that vvols in the vvol hierarchy are not orphaned in delete and revert use cases. Differences between vvols in the COW-based vvol hierarchy are used to clone the hierarchy and to migrate the hierarchy to a redo-based disk hierarchy.