Data-less Writes for Instantaneous Storage Restore
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional storage systems face inefficiencies and resource wastage during restore operations, as they require reading and writing data from snapshots, which can be time-consuming and consume system resources, especially when numerous data writes have occurred since the last snapshot.
Innovation Solution
The implementation of data-less writes, where changed data extents in a logical object are identified and communicated through the device stack without actual data payload, allowing for instantaneous restore operations by swapping file IDs and using a snapback driver to coalesce changes into larger regions, thereby reducing the need for replication and optimizing the restore process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional restore operations are performed by reading and writing data from snapshots, then data consistency is ensured, but restore time and system resource consumption increase significantly
Solution Approach 1:
The patent extracts only the essential metadata information (data extent changes) from the snapshot restore process, separating it from the actual data copying operation. By identifying and tracking only the data extents that have changed since the snapshot was taken, the system avoids reading and writing unnecessary unchanged data, thereby dramatically reducing restore time while maintaining data consistency through the extracted change information.
Solution Approach 2:
The system performs preliminary tracking of data extent changes between the snapshot creation time and the restore operation. By maintaining records of which data extents have been modified (through write operations) in the period between snapshot and restore, the system prepares advance knowledge of what needs to be restored, eliminating the need for full data re-copying and significantly accelerating the restore process.
2Reliability
If full data copying is performed during restore operations, then complete data restoration is achieved, but system resource utilization increases
Solution Approach 1:
The patent extracts and utilizes only the necessary information (data extent change metadata) from the snapshot structure, rather than copying the entire data set. By identifying which specific data extents need restoration through metadata analysis, the system restores only the minimal required data portions, ensuring complete restoration of changed data while minimizing system resource consumption during the restore operation.
3Reliability
If replication is halted during restore operations, then data consistency is maintained, but the window for halting replication increases system downtime
Solution Approach 1:
The system performs preliminary identification of changed data extents before the restore operation begins. By having advance knowledge of exactly which data extents require restoration (through prior tracking of write operations), the restore process can proceed immediately without needing to halt replication for extended periods, thereby maintaining data consistency while minimizing system downtime and improving availability.
Solution Approach 2:
The patent enables the restore operation to skip the time-consuming phases of reading and writing unchanged data by directly targeting only the identified changed data extents. This rushing through of the restore process using pre-identified change information significantly reduces the window during which replication must be halted, thereby maintaining data consistency while minimizing impact on system productivity and availability.
Data Source
AI summary
A method is used in managing restore operations in storage systems. A state of a logical object is retrieved and a second logical object is placed in a pre-step mode. An instantaneous restore operation between the first logical object and the second logical object is performed. Changed data extents in the first logical object are identified. Data-less writes are sent to any identified extents in the first logical object.