Distributed Storage I/O Continuity Across Site Link Failures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage systems with multiple sites face downtime due to temporary network disconnections, as they lack efficient mechanisms to maintain data coherence and allow I/O operations when links between sites fail, particularly in systems with more than three sites.
Innovation Solution
A distributed storage system with mirrored sites uses update records and bitmaps to track memory block changes, designating a primary site for continuous I/O operations during link failures, and synchronizing non-primary sites upon link restoration, utilizing primary version indicators to manage data consistency across multiple sites.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If coordination is required to ensure data coherence across multiple sites, then data integrity is maintained, but I/O operations must be stopped when network links fail, causing downtime
Solution Approach 1:
The system segments the multi-site storage network into isolated site groups when link failures occur. Each site or group of sites can independently process I/O operations within its segment, eliminating the need to stop all operations system-wide. This segmentation allows continuous I/O access at unaffected sites while maintaining data coherence through selective synchronization.
Solution Approach 2:
The patent introduces update records as intermediary data structures that track I/O operations and synchronization status across sites. These update records serve as mediators that enable sites to independently process operations and then reconcile data coherence when links are restored, without requiring continuous coordination during network failures.
2Productivity
If all sites are allowed to concurrently perform I/O operations to the same blocks, then I/O performance is improved, but coordination complexity increases to maintain coherence
Solution Approach 1:
Each storage site autonomously processes I/O operations and maintains its own update records without requiring real-time coordination with other sites. Sites independently track their own I/O operations and synchronization states, eliminating complex inter-site coordination mechanisms while still ensuring data coherence through the update record system.
Solution Approach 2:
The system uses update records as copies of I/O operation metadata that are maintained at each site. Instead of requiring sites to directly coordinate and share state, each site maintains a local copy of the synchronization information, allowing independent operation while preserving the ability to reconcile differences when needed.
3Reliability
If known solutions are used for two-site systems with coherence restoration, then data coherence is restored after link failures, but the solution is limited and cannot handle systems with more than three sites
Solution Approach 1:
The update record mechanism is designed as a universal solution that functions identically regardless of the number of sites in the system. The same update record data structures and synchronization protocols used in two-site systems are applied to multi-site configurations, allowing the system to scale from 2 to many sites without requiring different coordination mechanisms.
Solution Approach 2:
The patent extends the coordination problem from pairwise site relationships to multi-dimensional site groupings. By using update records that track I/O operations across arbitrary numbers of sites simultaneously, the system moves from solving a two-site problem to handling n-site configurations through the same fundamental mechanism.
Data Source
AI summary
Systems and methods for allowing and maintaining Input/Output (I/O) operations in a distributed storage system with multiple storage sites when communications links between the sites fail are provided. Mirrored storage sites use an update policy, including an update record and a bitmap configured to track changes in memory blocks storing an I/O request. When a link between two or more sites fails, one or more of the sites may be designated as a primary site, on which normal I/O operations may continue. One or more sites may also be designated as non-primary sites and are “detached” with respect to the primary site or sites. I/O operations on the non-primary sites may be suspended until the link is restored. When the link is restored, the update records and bitmaps are used to rebuild the detached sites.