Storage Data Allocation Using Orthogonal Latin Squares
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems face inefficiencies in data rebuilding speed and capacity utilization due to the heavy workload on surviving storage units during data reconstruction, especially as storage capacity increases, leading to prolonged data recovery times and vulnerability during array maintenance.
Innovation Solution
The proposed solution involves categorizing storage units into groups and using mutually orthogonal Latin squares to distribute data across storage units, allowing for parallel data reconstruction and reducing the workload on individual units, thereby increasing rebuilding speed and capacity efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data is stored using traditional storage allocation methods, then storage capacity is utilized, but data rebuilding speed is slow due to heavy workload on surviving storage units
Solution Approach 1:
The patent segments the storage array into multiple groups (e.g., Group 0, Group 1, etc.) where each group contains multiple storage units. This segmentation allows independent rebuilding operations within each group, reducing the workload on surviving units and enabling faster parallel reconstruction across groups.
Solution Approach 2:
The patent introduces a group dimension beyond individual storage units, organizing units into hierarchical groups. This dimensional change enables parallel rebuilding operations across multiple groups simultaneously, significantly improving rebuilding speed while maintaining data reliability through distributed redundancy.
2Reliability
If storage capacity is increased to improve data security, then data safety is enhanced, but rebuilding workload increases causing prolonged recovery times
Solution Approach 1:
By segmenting the storage array into multiple groups with independent rebuilding capabilities, the patent enables parallel reconstruction operations. This reduces the total reconstruction time even as storage capacity and redundancy increase, because multiple groups can rebuild simultaneously rather than sequentially.
Solution Approach 2:
The patent implements partial action by allowing rebuilding operations to proceed independently in each group rather than requiring the entire array to participate. This partial parallelism reduces overall reconstruction time while maintaining full data security through the distributed group structure.
3Productivity
If traditional data allocation is used, then storage is utilized, but capacity efficiency decreases due to redundant data distribution
Solution Approach 1:
The patent applies local quality by allocating redundant data differently across groups rather than uniformly across all units. Each group has optimized redundancy characteristics tailored to its specific configuration, improving overall capacity efficiency while maintaining data security through localized redundancy strategies.
Data Source
AI summary
An apparatus including a control unit, a memory having computer program code, and N groups of storage units electrically connected to the control unit is disclosed. Each of the N groups of storage units has N storage units, each of the N storage units has N storage regions, wherein N is a positive integer. The memory and the computer program code configured to, with the control unit, cause the apparatus to perform: storing a first data segment into an ith storage region of a first storage unit of a kth group of storage units; storing a fourth data segment into an ith storage region of a first storage it of a (k+1)th group of storage units; storing a fifth data segment into an ith storage region of a second storage unit of the (k+1)th group of storage units; and storing a sixth data segment into an ith storage region of a third storage unit of the (k+1)th group of storage units. Wherein the first data segment is associated with the fourth data segment, the first data segment is independent of the fifth data segment, and the first data segment is independent of the sixth data segment.