Journal-Based Data Replication Failover Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data protection systems face challenges in minimizing downtime during recovery and keeping pace with high data transaction rates at production sites, leading to potential production site shutdowns due to backlog of un-logged transactions at backup sites.
Innovation Solution
Implementing a method where the second site receives I/O from the first site, stores them in a journal, and regulates acknowledgement delays and journal size by controlling the response of a splitter to manage data lag and reduce the amount of IOs to be processed during failover, enabling instant access and reducing failover time to less than a couple of seconds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the backup site processes all data transactions synchronously with the production site, then data consistency is maintained, but the production site must slow down and may shut down due to high overhead
Solution Approach 1:
The patent segments the data transaction processing into two independent paths: a fast asynchronous path that maintains data consistency through journaling without blocking production transactions, and a slower synchronous path that applies journal entries to the backup site. This segmentation allows the production site to maintain high transaction rates while the backup site processes data at its own pace.
Solution Approach 2:
The patent introduces a journal as an intermediary data structure that decouples the production site from the backup site. The journal captures data transactions asynchronously and serves as a buffer, allowing the production site to continue processing transactions without waiting for the backup site to catch up, thereby maintaining both data consistency and high productivity.
2Loss of time
If the backup site processes data transactions quickly, then failover time is reduced, but data lag between sites increases causing loss of recent data
Solution Approach 1:
The patent applies preliminary action by pre-processing journal entries and maintaining a ready-to-apply buffer at the backup site. When failover occurs, the backup site can immediately continue processing from the last applied journal entry without needing to reprocess entire transaction histories, thereby minimizing failover time while preventing data loss through the continuous journal buffer.
3Productivity
If the system allows high data transaction rates at the production site, then productivity is maintained, but the backup site cannot keep pace leading to transaction backlog and potential shutdowns
Solution Approach 1:
The patent implements dynamics by making the backup site's processing rate adaptive rather than fixed. The backup site dynamically adjusts its processing speed based on the incoming journal entry rate and its current processing capacity, using flow control mechanisms to prevent backlog while ensuring all transactions are eventually processed, thereby maintaining reliability at varying production rates.
Data Source
AI summary
A method, system, and computer product for data replication comprising enabling a second site to receive I/O from a first site; wherein the second site stores the I/O in a journal until transferring the I/O to an image, regulating, at the first site, an application by delaying acknowledgements of I/O requests from the application, and controlling the size of a do stream in the journal by slowing a response of a splitter.