Disk Access Pattern Selection for Content-Based Storage RAID
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Storage systems face challenges in maintaining pace with high write rates at production sites without causing queuing and latency issues at backup sites, particularly due to the overhead of journaling and concurrent disk access patterns.
Innovation Solution
A method is implemented to select the order of accessing disks using a finite field-based disk access sequence, ensuring each entity accesses a predetermined number of disks uniquely, thereby minimizing simultaneous access to the same disk and reducing queuing and latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple consecutive accesses are made to the same disk, then data can be written efficiently, but large queues form causing high disk latency and I/O latency
Solution Approach 1:
The patent segments the disk access operations by distributing multiple consecutive accesses across different disks in the array. Instead of concentrating all accesses on a single disk, the system divides the access pattern across multiple disks, thereby preventing queue formation on any individual disk while maintaining overall write efficiency.
Solution Approach 2:
The patent introduces a new dimension to disk access by using multiple disks in the array simultaneously. Rather than accessing disks sequentially in one dimension, the system distributes accesses across multiple disks, effectively adding a spatial dimension to the access pattern and reducing temporal delays.
2Productivity
If multiple entities access disks concurrently, then system throughput increases, but simultaneous access to the same disk causes queuing and latency
Solution Approach 1:
The patent segments the access requests from multiple entities and distributes them across different disks in the array. By dividing the access workload and assigning different entities to different disks, the system maintains high throughput while avoiding simultaneous access conflicts that would cause queuing and latency.
Solution Approach 2:
The patent implements dynamic disk selection where the system adaptively assigns disks to different entities based on current system state. This dynamic allocation allows multiple entities to access disks concurrently without fixed assignment, optimizing throughput while preventing access conflicts through flexible resource distribution.
Data Source
AI summary
A disk access request is received at a storage system comprising a plurality of disks, from an entity having a substantially unique identifier. A finite field is defined, corresponding to the plurality of disks and based on the storage system array access width. A substantially unique sequence interval is computed, associated with the finite field and based on the substantially unique identifier and the array access width. A respective disk access sequence is generated, defining a disk access pattern the entity to access all disks in the array, wherein the sequence directs the entity to access at least the disks in a predetermined order that increments from a starting point to subsequent disks in accordance with the sequence interval, such that, when the disk access sequence is complete, the respective entity has accessed each disk in the portion of the plurality of disks only a predetermined number of times.