Journal-Based Storage Recovery from Cache Failure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data protection systems face inefficiencies in recovering from cache loss events, requiring the re-reading of entire replica storage data, which is time-consuming and unnecessary, especially in synchronous replication scenarios where no data loss occurs.
Innovation Solution
A journal-based data protection system that splits write I/Os to a replica site, stores them in a cache, periodically de-stages data to a volume, and restores lost write I/Os from the cache using the split write I/Os to the replica site, allowing for quick recovery without re-applying existing data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional data protection systems re-read entire replica storage data after cache loss, then data recovery is achieved, but recovery time is excessively long and unnecessary data is re-read
Solution Approach 1:
The patent segments the recovery process into two distinct parts: (1) re-reading only the cache-sized data that may have been lost, and (2) utilizing the journal that already contains the split write I/Os for direct restoration. This segmentation eliminates the need to re-read the entire replica storage data, significantly reducing recovery time while maintaining data integrity.
Solution Approach 2:
The system performs preliminary actions by maintaining a journal that continuously records split write I/Os before any cache loss occurs. This pre-prepared journal contains all necessary information for recovery, allowing the system to quickly restore data without needing to re-process entire data sets after a failure event.
2Reliability
If conventional systems re-apply existing data during recovery, then data consistency is restored, but the process is time-consuming and inefficient
Solution Approach 1:
The patent uses the journal as a copy of the write I/O history that can be directly applied during recovery. Instead of re-reading and re-processing entire data sets from storage, the system copies the necessary write operations from the pre-maintained journal, significantly improving recovery efficiency while ensuring data consistency.
Solution Approach 2:
The system discards the inefficient approach of re-reading entire replica storage data and recovers only the essential cache-sized data combined with the journal information. This selective recovery approach maintains data consistency while dramatically improving productivity during recovery operations.
3Reliability
If synchronous replication is used, then no data loss occurs, but conventional recovery still requires time-consuming re-reading of replica storage
Solution Approach 1:
The patent extracts only the necessary cache-sized data and journal information from the entire replica storage system for recovery operations. Even in synchronous replication where no data is lost, this extraction approach allows the system to bypass the time-consuming process of re-reading entire data sets, reducing recovery time while maintaining the data loss prevention benefits of synchronous replication.
Data Source
AI summary
In one aspect, a method includes receiving write I/Os for storage in a primary storage array, splitting the write I/Os to a replica site, storing the write I/Os for storage in the storage array in a cache, periodically de-staging the write I/Os in the cache to a volume and after a cache data loss, restoring lost write I/Os from the cache by using the write I/Os split to the replica site.