Storage Subsystem Consistency via Time-Synchronized Clocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data replication methods, such as IBM's Global Mirror, face challenges in maintaining consistency across storage volumes due to asynchronous data replication techniques, leading to potential out-of-order I/O operations and data integrity issues, which can result in inconsistent states and impact application performance.
Innovation Solution
Implementing a time-consistent methodology that synchronizes internal storage control clocks with a common clock source, using protocols like Network Time Protocol, to manage dependent write I/O operations and build consistency groups, ensuring that all storage subsystems coordinate and synchronize their times with a granularity smaller than the time of an I/O sequence, thus avoiding data freeze techniques and maintaining consistent states across volumes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If asynchronous data replication techniques are used to copy data between storage volumes, then data replication speed is improved, but data consistency across volumes deteriorates
Solution Approach 1:
The system performs preliminary actions by capturing write operations in a write buffer before they are committed to storage volumes. This allows the system to replicate data asynchronously while maintaining consistency by replaying captured writes in the correct sequence, resolving the contradiction between fast replication and data consistency.
Solution Approach 2:
The patent introduces intermediary components including a write buffer that captures write operations, a global sequence number generator that orders operations, and a consistency group manager that coordinates replication. These intermediaries enable asynchronous replication while ensuring consistent state across volumes.
2Reliability
If data freeze techniques are used to maintain consistency during replication, then data consistency is improved, but application performance deteriorates
Solution Approach 1:
Instead of freezing data during replication, the system performs preliminary actions by capturing write operations in a buffer before replication occurs. This allows continuous application access while maintaining consistency through ordered replay of captured writes, eliminating the need for data freeze.
Solution Approach 2:
The patent uses copying mechanisms to replicate data states without freezing the source. By capturing writes in a buffer and replaying them on target volumes, the system creates consistent copies while the source continues to accept writes, maintaining application performance.
3Productivity
If out-of-order dependent writes are allowed in asynchronous replication, then replication efficiency is improved, but data integrity deteriorates
Solution Approach 1:
The system performs preliminary ordering of write operations by capturing them in sequence in a write buffer and assigning global sequence numbers. This preliminary action allows efficient asynchronous replication while guaranteeing data integrity through ordered replay, preventing out-of-order writes.
Solution Approach 2:
The patent implements feedback mechanisms where the consistency group manager monitors replication status and uses sequence numbers to ensure correct ordering. This feedback loop maintains data integrity by detecting and correcting any out-of-order write scenarios while preserving replication efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided are a method, system, and article of manufacture, wherein a primary control unit sends a selected time to a plurality of secondary control units. Clocks are periodically synchronized at the primary control unit and the secondary control units with a time server. The primary control unit and the plurality of secondary control units consistently update secondary storage subsystems couplable to the plurality of secondary control units at the selected time.