Storage System Migration Without Host I/O Interruption
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems in Active-Active HA configuration face challenges in performing data migration without stopping host I/Os, leading to a loss of high availability when trying to maintain the HA configuration.
Innovation Solution
The solution involves a storage system with multiple storage apparatuses that migrate data from a pre-migration volume group to a post-migration volume group, where write commands are processed in a sequential order, and exclusive locks are obtained and released by storage apparatuses to ensure continuous host I/O operations during migration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data migration is performed in existing storage systems, then data can be moved from pre-migration volume group to post-migration volume group, but host I/O operations must be stopped and HA configuration must be canceled, resulting in loss of high availability
Solution Approach 1:
The patent segments the write command processing into multiple sequential steps across different storage apparatuses. Each storage apparatus processes write commands independently in a predetermined order, allowing continuous I/O operations while migration proceeds. This segmentation enables both high availability and data migration capability to coexist.
Solution Approach 2:
The patent introduces an intermediary locking mechanism that coordinates write operations across multiple storage apparatuses without stopping host I/Os. The locking mechanism acts as a mediator that ensures data consistency during migration while maintaining system availability, resolving the contradiction between migration and high availability.
2Productivity
If write commands are processed simultaneously across multiple storage apparatuses, then I/O speed is maintained, but data consistency cannot be ensured during migration
Solution Approach 1:
The patent implements preliminary actions by establishing a predetermined processing order for write commands before migration begins. Each storage apparatus knows in advance its position in the sequence, allowing simultaneous processing that maintains both speed and consistency. The first storage apparatus in the sequence writes without locking, while others obtain exclusive locks before writing.
Solution Approach 2:
The patent changes the parameter of write command processing from simultaneous to sequential with predetermined ordering. By modifying the processing parameter (order of execution) rather than the speed parameter, the system maintains high I/O throughput while ensuring data consistency through the structured sequence of operations.
3Manufacturing precision
If exclusive locks are obtained by all storage apparatuses during write operations, then data consistency is maintained, but I/O performance degrades due to serialization
Solution Approach 1:
The patent applies asymmetry by treating the first storage apparatus in the predetermined order differently from the others. The first apparatus writes without obtaining exclusive locks, while subsequent apparatuses obtain locks before writing. This asymmetric approach reduces serialization overhead while maintaining data consistency, preserving I/O performance.
Solution Approach 2:
The patent introduces dynamic behavior where storage apparatuses adapt their locking behavior based on their position in the predetermined sequence. The locking mechanism is dynamically applied only where necessary (to apparatuses other than the first), optimizing the balance between consistency and performance during migration operations.
Data Source
AI summary
Provided is a control method for a plurality of storage apparatuses that perform a migration from a pre-migration volume group to a post-migration volume group. The plurality of storage apparatuses are configured to accept a write command directed to any one of a plurality of volumes in the pre-migration volume group and the post-migration volume group when the migration is being processed. The control method includes writing write data of the write command to all of the plurality of volumes in a predetermined order, writing the write data to a last volume last among the plurality of volumes, and writing the write data to volumes different from the last volume after individually obtaining exclusive locks and, after completion of the writing the write data to the last volume, releasing the exclusive locks.