Mapped RAID Group Configurable Width Parallel Rebuild
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage systems face challenges with RAID data protection, including increased rebuild times due to write bandwidth bottlenecks and the risk of data loss from double disk failures during rebuilds, and lack support for RAID-10 configurations.
Innovation Solution
Implementing a mapped RAID group approach that divides physical drives into drive extents, allocates these extents from a pool to form mirrored sets, and performs logical block rotation to distribute data across multiple drives, allowing for configurable RAID widths and concurrent processing without a separate striping layer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is rebuilt to a single dedicated spare drive, then data protection is maintained, but write bandwidth bottleneck increases rebuild time
Solution Approach 1:
The patent segments the rebuild process by dividing the failed drive's data into multiple chunks that are simultaneously rebuilt across multiple spare drives in parallel, rather than sequentially rebuilding to a single spare drive. This segmentation of the rebuild workload eliminates the write bandwidth bottleneck and significantly reduces rebuild time while maintaining data protection through redundant copying across multiple drives.
Solution Approach 2:
The patent merges multiple spare drives into a collective rebuild target pool, allowing the system to simultaneously write reconstructed data to multiple drives in parallel. This combining of multiple write operations into concurrent operations increases the aggregate write bandwidth available for rebuild operations, resolving the bottleneck that limits single-drive rebuild performance.
2Reliability
If a dedicated hot spare disk is used for each RAID group, then data protection is provided, but write bandwidth of the spare disk becomes a performance bottleneck
Solution Approach 1:
The patent segments the rebuild workload by dividing it into multiple parallel reconstruction tasks that can be executed simultaneously across multiple spare drives, transforming a single-threaded rebuild process into a multi-threaded operation that fully utilizes available write bandwidth from multiple drives.
Solution Approach 2:
The patent makes multiple spare drives universally available as a shared rebuild pool for any failed drive in the RAID array, rather than dedicating a single spare to each RAID group. This universal spare pool allows any combination of failures to be handled by drawing from the collective spare resources, improving both productivity and resource utilization.
3Productivity
If data is striped across multiple drives to improve performance, then concurrent processing increases, but complexity of RAID configuration increases
Solution Approach 1:
The patent implements dynamic RAID width configuration that allows the striping factor to be adjusted based on workload requirements and hardware capabilities. The system can dynamically optimize the number of drives participating in the stripe, balancing performance gains against complexity management, and adapting to different failure scenarios without requiring fixed complex configurations.
Data Source
AI summary
A mapped RAID group is generated that contains multiple RAID extents, and each one of the RAID extents is configured and arranged to indicate multiple drive extents that are each a contiguous region of non-volatile data storage in a physical drive. The RAID extents are divided into equal size subsets based on a requested RAID width. Drive extents are allocated from a drive extent pool such that the drive extents indicated by RAID extents contained in the same subset are all allocated from different ones of the physical drives. Logical block rotation is performed between the storage object and the mapped RAID group, causing host data written to consecutive blocks of the storage object address space to be directed to different ones of the RAID extents in the mapped RAID group. Data mirroring is performed between drive extents contained within each of the individual RAID extents of the RAID group.