Cluster Volume Backup via Proxy Node Snapshot Provider
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed computing environments, existing backup solutions for cluster volumes are serialized, leading to heightened network traffic and degraded IO performance due to the requirement that only one node can create a backup at a time, as all I/O operations must be forwarded to the owner node for data consistency.
Innovation Solution
Implementing a snapshot provider that allows any node in a cluster to back up a volume using a proxy cluster volume, enabling point-in-time imaging and copy-on-write algorithms to create snapshots on a different node than the owner, while maintaining data consistency and allowing non-owner nodes to perform direct I/O operations, thus reducing the burden on the owner node.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all I/O operations are forwarded to the owner node for data consistency during backup, then data consistency is ensured, but network traffic increases and IO performance degrades
Solution Approach 1:
The patent segments I/O operations into two types: metadata operations (which require owner node coordination for consistency) and data operations (which can be performed locally by any node). This segmentation allows non-owner nodes to perform direct I/O to the cluster volume for data operations, reducing network traffic while maintaining consistency for metadata operations through the owner node.
Solution Approach 2:
The patent introduces a proxy cluster volume as an intermediary layer between non-owner nodes and the actual cluster volume. The proxy volume allows non-owner nodes to perform I/O operations without directly accessing the owner node, acting as a local cache that reduces network traffic while maintaining data consistency through coordination with the owner node for metadata changes.
2Reliability
If only one node is designated as owner for backup operations, then data consistency is maintained, but backup operations become serialized and IO performance degrades
Solution Approach 1:
The patent enables multiple nodes to perform backup operations simultaneously by making the backup capability universal across all cluster nodes. Non-owner nodes can initiate and perform backup operations directly on the cluster volume without being serialized through the owner node, while the owner node continues to manage metadata consistency. This multi-functionality allows parallel backup operations across multiple nodes, significantly improving backup throughput.
3Reliability
If non-owner nodes cannot perform direct I/O to cluster volume, then data consistency is simplified, but IO performance and accessibility are degraded
Solution Approach 1:
The patent implements local quality by allowing non-owner nodes to perform direct I/O operations locally on the cluster volume for data operations, while the owner node maintains global quality by managing metadata consistency. This differentiated approach enables high-speed local I/O operations at each node while maintaining overall data consistency through the owner node's coordination of metadata changes.
Data Source
AI summary
Implementing point in time back-up. A method that may be practiced in a distributed computing environment. The distributed computing environment includes a number of computing system nodes. The method includes at a first node in the computing environment, taking a snapshot of a cluster volume. The cluster volume stores data items (e.g. files/blocks). Taking a snapshot includes marking a point in time for data items in the volume. The method further includes receiving a notification that a different node in the computing environment than the first node, has determined to change a data item in the cluster volume at a point after the marked point in time from an old version of the data item to a new version of the data item. The method further includes at the first node, saving the old version of the data item prior to modification.