Cluster Data Synchronization via Non-Blocking Copy Sweep
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In active-active clustered systems, synchronization of data between source and target storage devices is challenging due to conflicts from concurrent I/O operations, especially when data is frequently changing, leading to synchronization failures and delays.
Innovation Solution
A method where a first computing device performs a copy sweep operation on a source storage device, determines if it has failed, and sends a message to a second computing device to suspend I/O operations without waiting for acknowledgement, allowing the copy sweep operation to be retried without delays, and resumes I/O operations based on completion or predetermined time periods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the first computing device waits for acknowledgement from the second computing device before retrying the copy sweep operation, then communication reliability is improved, but synchronization speed deteriorates
Solution Approach 1:
The first computing device sends the suspend message to the second computing device before retrying the copy sweep operation, without waiting for an acknowledgement. This preliminary action eliminates the waiting time and allows the retry to proceed immediately, thus improving synchronization speed while maintaining communication reliability through the eventual execution of the suspend message
Solution Approach 2:
The patent skips the acknowledgement reception step in the communication flow. The first computing device rushes through the message sending process by not pausing to receive confirmation, thereby eliminating communication delays and accelerating the copy sweep retry process
2Productivity
If the copy sweep operation is retried immediately without receiving acknowledgement, then synchronization speed is improved, but communication reliability deteriorates
Solution Approach 1:
The suspend message is sent in advance as a preliminary action before the retry is attempted. This ensures that the second computing device is already notified and will suspend I/O operations when the message is processed, maintaining reliability even though the retry proceeds without waiting for explicit confirmation
3Reliability
If acknowledgement messages are required for every copy sweep retry, then communication reliability is improved, but communication overhead increases
Solution Approach 1:
The patent extracts and removes the acknowledgement message requirement from the copy sweep retry process. By taking out this communication step, the system reduces communication overhead and message traffic while still achieving the necessary coordination through the unilateral suspend message mechanism
Data Source
AI summary
A method, computer program product, and computer system for performing, via a first computing device, a copy sweep operation to a first range of data on a source storage device. It may be determined that the copy sweep operation has failed. A message may be sent to a second computing device to suspend I/O operations to the first range of data. The copy sweep operation may be retried based upon, at least in part, determining that the copy sweep operation has failed, wherein the copy sweep operation may be retried without the first computing device receiving acknowledgement that the second computing device is suspending the I/O operations to the first range of data.