Storage Node Redundancy During Maintenance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In Software Defined Storage (SDS) systems, maintaining data redundancy during node maintenance is time-consuming and resource-intensive, leading to reduced I/O performance, especially when multiple nodes require sequential maintenance, resulting in potential data loss if one node fails during the process.
Innovation Solution
A computer system with multiple nodes and a management computer that forms High Availability (HA) pairs, where write data is duplicated between two nodes and also written to a third node, allowing for efficient data redundancy maintenance by creating recovery volumes and journal logs to ensure data integrity and reduce downtime.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is completely copied from the maintenance target node to another node before maintenance, then data redundancy is maintained, but time and resources are significantly consumed
Solution Approach 1:
The patent creates a recovery volume in advance on a third node before maintenance begins. This recovery volume is prepared as a standby destination, so when maintenance occurs and the second node becomes unavailable, data can be immediately redirected to the pre-prepared recovery volume without needing to create it during the maintenance window, thus reducing the time penalty while ensuring redundancy is maintained
Solution Approach 2:
The patent introduces a third node as an intermediary recovery volume destination. Instead of directly copying data between the first and second nodes, the third node serves as a mediator that receives data during maintenance periods. This intermediary approach allows data to be preserved with redundancy even when the primary HA pair is temporarily disrupted during node maintenance
2Reliability
If data is completely copied from the maintenance target node to another node before maintenance, then data redundancy is maintained, but resources are significantly consumed
Solution Approach 1:
The patent applies partial action by only copying data to the recovery volume when absolutely necessary (i.e., when the second node becomes unavailable during maintenance). During normal operation, the HA pair continues to function without triggering full data copying. This selective approach reduces resource consumption compared to continuously maintaining full redundancy, while still ensuring data safety when needed
3Productivity
If maintenance is performed on a node-by-node basis while the cluster is in operation, then continuous service is maintained, but data loss occurs if a failure is occurred during the process
Solution Approach 1:
The patent prepares a recovery volume in advance on a third node before maintenance begins. This pre-prepared recovery volume acts as a cushion or safety buffer that is already in place and ready to receive data if needed. When maintenance occurs and the second node fails, this pre-positioned recovery volume immediately provides the necessary data protection, preventing data loss without interrupting service to the first node
Data Source
AI summary
In a computer system 10 that includes multiple nodes 100 capable of storing data, and a management computer 410 that manages the nodes 100, a first volume of a first node and a second volume of a second node 100 constitute an HA (High Availability) pair for managing identical data in a duplicated manner. the CPU 120 of the second node is configured to, When the first node becomes offline, write data that is to be written into the second volume of the second node thereafter is written into the second volume and into a third volume of a third node that is different from the first node and the second node.