VM-Granularity Data Replication via Metadata Mirroring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data protection systems face challenges in ensuring robust and efficient data replication and failover between production and backup sites, particularly in maintaining data integrity and minimizing lag, especially in virtual environments.
Innovation Solution
The proposed solution involves intercepting I/O operations at a virtual machine, sending metadata and data through a data protection appliance, and ensuring metadata mirroring is complete before data mirroring occurs, using a splitter in the hypervisor to manage replication in a per-VMDK granularity, and utilizing service layers to track and synchronize changes for failover and recovery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data replication is performed at the data block level or file system level, then replication efficiency is improved, but data integrity and consistency between metadata and data are compromised
Solution Approach 1:
The patent segments the data protection system into distinct functional layers: a service layer that handles metadata operations and a data layer that handles actual data blocks. This segmentation allows independent optimization of each layer - the service layer can process metadata efficiently while the data layer maintains integrity through coordinated replication. The virtual machine disk is also segmented into metadata structures and data blocks, enabling granular control over replication priorities.
Solution Approach 2:
The patent implements preliminary action by requiring that metadata replication be completed and acknowledged before data replication begins. The service layer first replicates the metadata structure to the remote site and receives confirmation of successful metadata establishment. Only after this preliminary metadata setup is complete does the system proceed with data block replication, ensuring that the destination site has the proper structural framework in place to receive and organize incoming data, thereby preventing data integrity issues.
2Reliability
If metadata is replicated before data, then data integrity is improved, but replication time increases
Solution Approach 1:
The patent maintains continuity of useful action by implementing an overlapping replication strategy. While metadata replication is completing at the service layer, the system prepares data blocks for replication and pre-configures the destination storage structures. Once metadata replication is acknowledged, data replication immediately begins without idle gaps. The system continuously processes I/O operations, maintaining replication streams active whenever possible, thereby minimizing total replication time while preserving the critical metadata-first sequence for integrity.
3Manufacturing precision
If per-VMDK granularity replication is implemented, then data protection precision is improved, but system complexity increases
Solution Approach 1:
The patent implements universality by creating a multi-functional service layer that handles multiple types of operations through a unified interface. The service layer can process metadata replication, data block replication, I/O interception, and coordination tasks through the same fundamental mechanisms. This multi-functionality allows per-VMDK granularity without proportionally increasing complexity, as the same service layer infrastructure serves multiple purposes across different virtual machine disks and replication scenarios.
Data Source
AI summary
In one aspect, a method includes intercepting an I/O sent by a virtual machine at a first site, sending the I/O to a first data protection appliance (DPA) at the first site, sending I/O metadata to a first service layer at the first site, sending I/O data down the I/O stack after receiving acknowledgement that the I/O metadata arrived at the first service layer, intercepting the I/O at the first service layer, sending the I/O metadata from the first service layer to a second service layer at a second site and sending I/O data to the second service layer at the second site by ensuring that I/O metadata mirroring between the first site and the second site is completed before I/O data mirroring occurs between the first site and the second site.