RAID Rebuild Acceleration via External Mirrored Copy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As non-volatile data storage systems grow in density, the rebuild time for RAID arrays becomes increasingly lengthy, especially under load conditions, due to conflicts between rebuild activities and user workload, which can be exacerbated by the size of the arrays and the type of storage devices used, such as SSDs and HDDs.
Innovation Solution
The implementation of mirrored RAID arrays that utilize global intelligence to manage rebuild processes by transferring data between arrays over a high-performance computing fabric, allowing for direct memory access (DMA) or remote DMA (RDMA) to accelerate the rebuild process, and redirecting user requests to minimize load on the rebuilding array.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional on-array rebuild is used, then data integrity is maintained, but rebuild time becomes excessively long
Solution Approach 1:
The patent utilizes mirrored copies of data stored on a second RAID array to accelerate the rebuild process. Instead of reconstructing data from scratch using parity information, the system copies data directly from the mirror array, significantly reducing rebuild time while maintaining data integrity through the redundant mirror copy.
Solution Approach 2:
The system introduces a second RAID array as an intermediary that holds mirrored copies of data. This intermediary array enables fast data transfer during rebuild operations, acting as a mediator between the failed drive and the replacement drive, thereby reducing the time required for data reconstruction.
2Reliability
If rebuild operations proceed under load conditions, then system availability is maintained, but rebuild performance deteriorates
Solution Approach 1:
The system performs preliminary actions by maintaining mirrored copies of data on a second RAID array before failures occur. This pre-positioned data allows the rebuild operation to proceed rapidly without waiting for data reconstruction, enabling the system to maintain both availability and performance during rebuild operations.
Solution Approach 2:
By using mirrored copies stored in advance on a second array, the system enables fast data transfer during rebuild operations even when under load. The pre-existing mirrors eliminate the need for real-time data reconstruction, allowing rebuild to proceed at high speed without sacrificing system availability.
3Reliability
If data transfer is performed sequentially, then data integrity is ensured, but transfer speed is limited
Solution Approach 1:
The patent segments the data transfer process by utilizing multiple drives in parallel within the RAID arrays. Instead of transferring data sequentially from a single source drive, the system reads from multiple drives simultaneously and writes to multiple drives in parallel, significantly increasing transfer speed while maintaining data integrity through RAID's distributed redundant storage.
Solution Approach 2:
The mirrored copy structure enables parallel data transfer operations. Both arrays can read and write operations proceed simultaneously during the rebuild process, with the second array providing fast read access and the first array performing write operations to the replacement drive, thereby increasing overall transfer speed while ensuring data integrity.
Data Source
AI summary
When rebuilding a RAID (Redundant Array of Independent Disks) array in which a drive has failed, if another RAID array contains a mirror copy of the of the rebuilding RAID array content, this mirroring RAID array can be used to more rapidly rebuild the RAID array with the failed drive. Data requests to the rebuilding RAID array can be redirected to the mirroring RAID array; data can be transferred from the mirroring RAID array; or a combination of these can be used to finish rebuilding more quickly. When transferring data to the rebuilding array from the mirroring array, the transfer can be performed as a direct memory access (DMA) process independently of the RAID module of either array.