Locality-Based Quorum Eligibility for Distributed Data Store Failover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed data stores employing single-master replication models face challenges in ensuring data durability and integrity when the master data store fails, as they require a consensus among all nodes for failover, which can be blocked by unresponsive or failed nodes.
Innovation Solution
Implementing a locality-based failover quorum that excludes failover-quorum-ineligible nodes, allowing the system to proceed with failover without waiting for unresponsive nodes, ensuring the newly elected master is aware of all committed updates and maintaining data durability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a consensus among all nodes is required for failover, then data durability and integrity are ensured, but failover speed is reduced and system availability deteriorates due to blocking by unresponsive nodes
Solution Approach 1:
The patent extracts problematic nodes from the failover quorum by identifying and excluding failover-quorum-ineligible nodes. This allows the system to proceed with failover using only eligible nodes, preventing blocking by unresponsive nodes while maintaining data durability through careful selection of nodes that have acknowledged committed updates.
Solution Approach 2:
The patent segments the node population into two distinct groups: failover-quorum-eligible nodes and failover-quorum-ineligible nodes. This segmentation allows the system to form a failover quorum from only the eligible subset, enabling faster failover while still ensuring data durability by selecting nodes that have properly acknowledged committed updates.
2Reliability
If the system waits for all nodes to respond before failover, then complete consensus is achieved, but system availability and responsiveness deteriorate due to prolonged delays
Solution Approach 1:
The patent introduces dynamic node eligibility determination where nodes are evaluated against specific criteria (acknowledgment of committed updates, current operational status) to determine their eligibility for the failover quorum. This dynamic approach allows the system to adaptively select eligible nodes rather than requiring static participation from all nodes, improving system availability while maintaining consensus accuracy.
Solution Approach 2:
The patent performs preliminary evaluation of node eligibility before initiating failover by checking whether nodes have acknowledged committed updates and are currently operational. This preliminary action identifies eligible nodes in advance, allowing the system to proceed with failover using only these pre-validated nodes, thereby improving system availability without sacrificing consensus accuracy.
3Adaptability or versatility
If unresponsive nodes are included in the failover quorum, then complete node participation is maintained, but failover reliability deteriorates due to arbitrary blocking delays
Solution Approach 1:
The patent applies different quality requirements to different nodes based on their individual states. Nodes are evaluated against local criteria (whether they have acknowledged committed updates, their current operational status) to determine eligibility. This local quality assessment ensures that only nodes with the appropriate state participate in the failover quorum, improving failover reliability while maintaining adaptability in node participation.
Data Source
AI summary
Disclosed are various embodiments for distributing data items. A plurality of nodes forms a distributed data store. A new master candidate is determined through an election among the plurality of nodes. Before performing a failover from a failed master to the new master candidate, a consensus is reached among a locality-based failover quorum of the nodes. The quorum excludes any of the nodes that are in a failover quorum ineligibility mode.