Active-Active Storage Node Transition via Invalid Track Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The transition to active-active mode in computer networks with replicated storage arrays is time-consuming due to the need for full consistency between replicas, which can take minutes, hours, or days, and existing methods do not allow for discovery and access while replicas are inconsistent.
Innovation Solution
An apparatus and method that enable discovery and access to replicas while they are inconsistent by converging characteristics and using access bias to resolve inconsistencies, with each storage array maintaining records of invalid extents and handling read/write commands to provide data based on validity and bias, allowing for preferential access and synchronization in the background.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If full consistency between replicas is enforced before enabling discovery and access, then data consistency is improved, but transition time to active-active mode increases
Solution Approach 1:
The system performs preliminary convergence of replica characteristics (metadata, configuration, states) before enabling discovery and access, while allowing data synchronization to continue in the background. This preliminary preparation enables the system to switch to active-active mode without waiting for complete data consistency, thereby reducing transition time while maintaining operational reliability.
Solution Approach 2:
The system dynamically manages replica consistency by allowing selective access to consistent portions of replicas while continuing synchronization for inconsistent portions. The consistency state is not static but evolves dynamically as data synchronizes in the background, enabling the system to transition to active-active mode progressively rather than waiting for complete consistency.
2Speed
If replicas are made accessible while inconsistent, then access time is improved, but data consistency deteriorates
Solution Approach 1:
The system applies different consistency requirements to different portions of the replica data. Consistent portions are made immediately accessible to hosts, while inconsistent portions continue to synchronize in the background. This local differentiation allows the system to provide fast access to available data while maintaining consistency guarantees where applicable.
Solution Approach 2:
The system implements partial consistency by allowing access to the extent that consistency has been achieved, rather than requiring complete consistency across all data. Hosts can access and operate on consistent data portions while the system continues to synchronize remaining portions, providing useful partial functionality immediately.
3Reliability
If complete synchronization is performed before active-active mode, then replica consistency is improved, but productivity during transition deteriorates
Solution Approach 1:
The system maintains continuous useful action by allowing discovery, access, and I/O operations to proceed while data synchronization continues in the background. Rather than halting productivity during the transition to active-active mode, the system enables both synchronization and operational activities to occur concurrently, maximizing system availability and productivity throughout the transition period.
Data Source
AI summary
In a data storage system in which a first storage array and a second storage array maintain first and second replicas of a production volume, the replicas are made discoverable and accessible while inconsistent. Each storage array maintains an invalid track list of inconsistencies. Initially, all tracks are marked as invalid. While background synchronization is eliminating inconsistencies, accesses to invalid tracks are resolved by exchanging data associated with IOs and updating the invalid track lists based on IO bias and other factors.