Virtual Cluster Intersite Data Node Failover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage systems face challenges in providing high availability across multiple sites due to high recovery time objectives, complex orchestration, and the need for additional components for failure detection and handling, which are not efficiently addressed by current disaster recovery solutions.
Innovation Solution
The creation of a virtual cluster spanning multiple data storage systems, where bidirectional remote replication is established between data nodes, allowing for failover processing and maintaining a single logical device view, ensuring that data nodes from different systems can read and write to the same virtual cluster logical device, and automatically replicate writes across sites.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If bidirectional remote replication is established between data nodes on different data storage systems, then intersite high availability is improved and recovery time objective is reduced, but device complexity increases due to the need to configure and manage replication relationships across multiple systems
Solution Approach 1:
The patent merges multiple data storage systems into a single virtual cluster, presenting them as one unified system to hosts. This consolidation simplifies management by reducing the number of discrete systems that need to be configured and monitored, while internally maintaining bidirectional replication between constituent systems for high availability.
Solution Approach 2:
The virtual cluster provides multi-functional capabilities by combining data storage, bidirectional replication, failure detection, and automated failover into a single unified architecture. This universal approach allows the system to perform multiple functions simultaneously without requiring separate dedicated components for each function.
2Reliability
If additional components are added for failure detection and handling in disaster recovery solutions, then reliability is improved, but device complexity and orchestration requirements increase
Solution Approach 1:
The data nodes within the virtual cluster perform self-monitoring for failures and automatically execute failover procedures without requiring external orchestration components. Each data node can detect its own status and trigger appropriate recovery actions, eliminating the need for complex centralized failure detection and handling infrastructure.
3Adaptability or versatility
If data nodes from different data storage systems access the same logical device, then data accessibility is improved, but the need for complex orchestration and additional components increases
Solution Approach 1:
The virtual cluster acts as an intermediary layer between data nodes from different storage systems and the logical devices they access. This mediator abstracts the underlying complexity of cross-system access, allowing data nodes to interact with unified logical devices while the virtual cluster manages the actual data routing and consistency across the distributed systems.
Data Source
AI summary
Creating and using a virtual cluster may include: creating a first cluster logical device on a first data storage system including data nodes; creating a second cluster logical device on a second data storage system including data nodes; configuring the first cluster logical device and the second cluster logical device as a same first logical device; establishing bidirectional remote replication between the first and second cluster logical devices; determining pairs of data nodes including a data node from the first data storage system and another data node from the second data storage system; determining a failure of a first data node on the first data storage system, wherein one of the pairs of data nodes includes the first data node and a second data node of the second data storage system; and responsive to determining the failure of the first data node, performing failover processing by the second data node.