Association-Based Disk Allocation for Even RAID Stripe Placement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems face inefficiencies in disk allocation, leading to prolonged allocation times, increased system overhead, and uneven distribution of RAID stripes, which compromises data recovery success rates due to the need to traverse all disks and nodes, resulting in some nodes and disks having more or fewer stripes.
Innovation Solution
A method for allocating storage space that selects storage nodes and disks based on association relationships, using neighborhood matrices to evenly distribute RAID stripes across nodes and disks, reducing allocation time and improving data recovery success.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all disks in the system are traversed for allocation, then complete disk allocation is achieved, but allocation time becomes too long
Solution Approach 1:
The patent segments the disk allocation process by dividing disks into different types (first type and second type) and allocating them through different paths. Management disks are allocated first, followed by data disks, avoiding the need to traverse all disks sequentially. This segmentation reduces allocation time while ensuring complete allocation of all disk types.
Solution Approach 2:
The patent performs preliminary allocation of management disks before allocating data disks. By pre-identifying and allocating management disks based on their specific requirements, the system prepares the foundation for subsequent data disk allocation, reducing the overall allocation time while ensuring complete allocation.
2Reliability
If all disks in the system are traversed for allocation, then complete disk allocation is achieved, but system overhead increases
Solution Approach 1:
The patent segments the allocation process into distinct phases for management disks and data disks, with each phase having specific allocation rules. This segmentation reduces system overhead by avoiding unnecessary traversal of all disks at each allocation stage, while still ensuring complete allocation of all disk types through the structured approach.
3Ease of manufacture
If traditional allocation methods are used, then allocation process is simple, but RAID stripe distribution becomes uneven
Solution Approach 1:
The patent applies different allocation strategies to different disk types based on their specific requirements. Management disks receive allocation according to their management functions, while data disks are allocated to achieve uniform RAID stripe distribution. This local quality approach ensures both process simplicity and distribution precision by tailoring allocation methods to local needs.
Solution Approach 2:
The patent changes the allocation parameters based on disk type, using different allocation rules for management disks versus data disks. For data disks, the system adjusts allocation parameters to ensure uniform RAID stripe distribution across the storage system, while maintaining simple allocation processes through clear parameter definitions.
4Loss of time
If disks are allocated without considering node associations, then allocation process is faster, but data recovery success rate decreases
Solution Approach 1:
The patent incorporates node association parameters into the allocation process by tracking and utilizing the relationships between storage nodes. The system uses parameters such as the number of stripes per node and association counts to make informed allocation decisions, ensuring that RAID stripes are distributed in a way that enhances data recovery success rate while maintaining efficient allocation timing.
Data Source
AI summary
Techniques for allocating storage space of disks in a storage system involve selecting a current storage node from a plurality of storage nodes according to at least one previously selected storage node and association relationships between the plurality of storage nodes. Such techniques further involve selecting a current disk from a plurality of disks according to at least one previously selected disk, the current storage node, and association relationships between the plurality of disks. Such techniques further involve allocating at least one slice in the current disk to a redundant array of independent disks (RAID) stripe. Such a technique can support a distributed storage system having more storage nodes, can greatly reduce the time for allocating the storage space of disks, and can make a RAID stripe evenly allocated between the storage nodes and the disks in the storage system, thereby greatly improving a success rate of data recovery.


