Clustered Storage Node Ownership Change via Segmented Virtual Volume Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In clustered storage arrays, changing ownership of virtual volume regions between nodes disrupts host access due to suspended IOs as metadata and logs are flushed across all virtual volumes simultaneously, causing significant disruption.
Innovation Solution
Increasing the number of zero logical disk regions and spreading data across them, allowing each node to own a set of zero LDs, and mapping virtual volumes to multiple groups of zero LDs, so only one group is affected during ownership changes, reducing the data to be flushed and minimizing IO disruption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If ownership of virtual volume regions is changed between nodes in a clustered storage array, then node reorganization and load balancing are achieved, but host IO access is disrupted because all virtual volumes must suspend IOs while metadata and logs are flushed simultaneously
Solution Approach 1:
The patent segments the virtual volume metadata and log structures into multiple independent regions (first region, second region, third region, fourth region) that can be flushed independently. This allows ownership changes to affect only specific regions rather than requiring simultaneous suspension of all virtual volume IOs, thereby maintaining host access continuity while enabling node reorganization.
Solution Approach 2:
The patent implements periodic or staged flushing of metadata and log regions rather than simultaneous flushing of all regions. The system flushes the first region and second region in one phase, then the third region and fourth region in another phase, allowing IOs to resume between phases. This periodic action reduces the duration of IO disruption while achieving complete ownership transfer.
2Reliability
If metadata and logs are flushed across all virtual volumes simultaneously during ownership changes, then data consistency is maintained, but the disruption to host IOs is maximized
Solution Approach 1:
The patent divides the metadata and log flushing operation into separate segments corresponding to different regions (first, second, third, fourth regions). Each region can be flushed independently and in parallel or sequential batches, reducing the total time IOs are suspended while maintaining data consistency through controlled region transitions.
Solution Approach 2:
The system performs preliminary actions by establishing the ownership structure and region mappings before initiating the flush operation. The metadata regions and log regions are pre-configured with their ownership relationships, allowing the flush to proceed efficiently through predetermined paths without requiring global IO suspension.
Data Source
AI summary
Each node in a clustered array is the owner of a set of zero logical disks (LDs). Thinly-provisioned VVs (TPVVs) are partitioned so each is mapped to a group of zero LDs from different sets of zero LDs. When there is a change in ownership, the affected zero LDs are switched one at a time so only a group of the TPVVs is affected each time.