Distributed Front-End RAID for Controller Bottleneck Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional RAID systems are inefficient due to dataflow bottlenecks, reliance on RAID controllers for data management, and complex partitioning schemes, which limit performance and reliability.
Innovation Solution
A front-end distributed RAID system that allows RAIDing on a per-data segment or per-file basis, eliminating the need for RAID controllers by distributing control functions among clients, a third-party device, or storage devices, and enabling direct data access through DMA or RDMA.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a RAID controller is used to manage all data storage operations, then data management and RAID functions are centralized, but dataflow bottlenecks occur and performance decreases
Solution Approach 1:
The patent divides the centralized RAID controller function into distributed segments across multiple storage devices. Each storage device independently manages its own data segments, eliminating the bottleneck of a single controller handling all data operations. This segmentation allows parallel data processing and removes the centralized control point that caused performance degradation.
Solution Approach 2:
The patent extracts RAID control functions from the centralized controller and assigns them to individual storage devices. By taking out the RAID management capability from the controller and embedding it in each storage device, the system eliminates the dataflow bottleneck while maintaining RAID functionality through distributed autonomous operation.
2Adaptability or versatility
If storage devices are partitioned for RAID access, then RAID groups can be formed, but access complexity increases and management becomes difficult
Solution Approach 1:
The patent implements self-service by allowing storage devices to autonomously identify themselves as part of RAID groups without requiring complex partitioning management. Each storage device independently determines its role and membership, eliminating the need for centralized partition management and reducing operational complexity while maintaining RAID adaptability.
3Reliability
If a dedicated region of storage is allocated for RAID, then RAID groups can be isolated, but space utilization is reduced and cannot be accessed independently
Solution Approach 1:
The patent segments storage space at the data segment level rather than creating dedicated regions. This allows RAID groups to be formed by segmenting data across multiple devices without reserving exclusive regions, thereby maintaining reliability through isolation while maximizing space utilization and enabling independent access to any storage space.
4Reliability
If RAID controllers are cross-connected for redundancy, then data availability is improved, but system cost and complexity increase dramatically
Solution Approach 1:
The patent extracts the redundancy function from the controller cross-connection architecture and implements it at the storage device level. Each storage device independently manages its own data and parity information, eliminating the need for complex controller cross-connections while maintaining data availability through distributed redundancy.
Solution Approach 2:
The patent implements self-service redundancy where each storage device autonomously manages its own data protection without requiring coordination through cross-connected controllers. This eliminates controller cross-connection complexity while maintaining reliability through independent device-level redundancy management.
Data Source
AI summary
An apparatus, system, and method are disclosed for a front-end, distributed redundant array of independent drives (“RAID”). A storage request receiver module receives a storage request to store object or file data in a set of autonomous storage devices forming a RAID group. The storage devices independently receive storage requests from a client over a network, and one or more of the storage devices are designated as parity-mirror storage devices for a stripe. The striping association module calculates a stripe pattern for the data. Each stripe includes N data segments, each associated with N storage devices. The parity-mirror association module associates a set of the N data segments with one or more parity-mirror storage devices. The storage request transmitter module transmits storage requests to each storage device. Each storage request is sufficient to store onto the storage device the associated data segments. The storage requests are substantially free of data.