RAID Storage Scaling via Drive Split Redistribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage systems that implement RAID face challenges in scaling storage capacity efficiently as commercially available drives increase in data storage capacity and cost, leading to costly increments of R+P drives that may not be utilized within a reasonable timeframe.
Innovation Solution
The solution involves organizing drives into fixed-size splits, implementing a selected RAID level, and scaling storage capacity by adding fewer than the initial number of new drives to a cluster, redistributing splits to create new protection groups, and potentially forming new clusters, allowing for more granular and flexible expansion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If storage capacity is scaled by adding R+P drives in traditional RAID, then storage capacity increases, but cost increases significantly and drives may not be utilized within a reasonable timeframe
Solution Approach 1:
The patent segments drives into multiple splits (e.g., dividing each drive into 2 or more splits). This allows the storage system to scale by adding individual drives and utilizing only the necessary splits, rather than requiring complete R+P sets. For example, if a drive is divided into 2 splits, adding one drive provides 2 splits that can be integrated into existing protection groups or used to create new ones, enabling granular capacity expansion without the overhead of adding entire R+P groups.
2Quantity of substance
If storage capacity is scaled by adding R+P drives, then storage capacity increases, but the system requires adding multiple drives at once
Solution Approach 1:
By dividing drives into splits, the system enables incremental scaling where individual drives can be added one at a time or in small batches. The controller dynamically allocates splits from new drives to existing or new protection groups, allowing flexible scaling operations that don't require coordinating additions of multiple drives simultaneously.
Solution Approach 2:
The patent implements dynamic split allocation where the controller can reassign splits between protection groups based on current system needs and available capacity. When new drives are added, their splits are dynamically integrated into the existing RAID structure, and protection groups can be dynamically recreated or expanded to utilize the new capacity efficiently.
3Reliability
If drives are organized into protection groups with fixed R+P structure, then data protection is maintained, but storage utilization efficiency decreases
Solution Approach 1:
The patent segments the traditional rigid protection group structure into finer-grained units (splits). Instead of requiring entire drives to be allocated to protection groups, individual splits can be allocated, allowing more efficient packing of storage capacity while maintaining the required redundancy. This enables higher utilization by reducing the overhead of unused capacity that occurs when drives are added in fixed R+P increments.
Solution Approach 2:
The system changes the fundamental parameter of protection group composition from whole drives to fractional splits. This allows the protection group size and composition to be dynamically adjusted based on actual capacity needs rather than fixed drive counts, optimizing both protection and utilization.
Data Source
AI summary
Each drive in a cluster is organized into splits of a single size. The splits are used as protection group members for a selected RAID level. Individual clusters can be scaled up with a number of new drives that is less than the number of protection group members for the RAID level by redistributing some data/parity to splits of the new drives. Splits are selected for redistribution such that new protection groups are created. If the number of new drives is equal to or greater than the number of protection group members for the RAID level, then new clusters may be created using the new drives. Any remaining drives are used to scale-up individual clusters.