Clustered RAID Data Organization with Slice-Based Extent Stores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Clustered storage systems face inefficiencies in parity overhead and distribution of storage containers among nodes, particularly in high availability configurations, leading to increased parity ratios and the need for additional RAID groups when nodes fail.
Innovation Solution
The organization of storage containers and RAID groups is optimized by forming RAID groups from slices of SSDs rather than entire SSDs, allowing multiple RAID groups to coexist on the same set of SSDs with varying parity ratios and enabling multi-host access, which facilitates even distribution of extent stores among nodes during failover.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If each storage container is stored in a separate RAID group to enable distribution among surviving nodes, then container distribution capability is improved, but parity overhead increases significantly
Solution Approach 1:
The patent segments storage containers into smaller units called extents, which can be distributed across multiple RAID groups. Instead of allocating one entire RAID group per container, the system divides containers into extent stores that can be spread across multiple RAID groups, thereby reducing the need for additional parity groups while maintaining distribution capability.
Solution Approach 2:
The patent introduces a new organizational dimension by creating extent stores that span multiple RAID groups. Rather than the traditional one-to-one mapping between containers and RAID groups, the system creates a many-to-many relationship where extent stores can be distributed across multiple RAID groups, effectively adding a dimensional layer to the storage organization hierarchy.
2Reliability
If additional RAID groups are created to house distributed containers from failed nodes, then high availability is improved, but device complexity increases
Solution Approach 1:
The patent makes RAID groups universal by allowing them to serve multiple purposes simultaneously. A single RAID group can contribute storage capacity to multiple different extent stores from different containers, rather than being dedicated to a single container. This multi-functionality reduces the total number of RAID groups needed while maintaining high availability through proper distribution.
Solution Approach 2:
The patent merges multiple container storage requirements into shared RAID groups. Instead of creating separate RAID groups for each container, the system combines storage needs from multiple containers into common RAID groups, where extents from different containers are distributed across the same physical storage resources, thereby reducing overall system complexity.
3Productivity
If SSDs are divided into multiple RAID groups with slices, then storage efficiency is improved, but device complexity increases
Solution Approach 1:
The patent segments SSD storage capacity into slices that can be allocated to multiple RAID groups simultaneously. Each SSD is divided into multiple slices, and these slices can be distributed across different RAID groups to serve multiple extent stores. This segmentation approach maximizes storage utilization while maintaining manageable complexity through standardized slice allocation.
Data Source
AI summary
In one embodiment, a node of a cluster having a plurality of nodes, executes a storage input/output (I/O) stack having a redundant array of independent disks (RAID) layer. The RAID layer organizes solid state drives (SSDs) within one or more storage arrays as a plurality of RAID groups associated with one or more extent stores. The RAID groups are formed from slices of storage spaces of the SSDs instead of entire storage spaces of the SSDs. This provides for RAID groups to co-exist on a same set of the SSDs.