Hierarchical Cluster Voting for Storage Node Resilience
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cluster voting systems in network attached storage devices are prone to failures, where a majority loss of nodes can lead to the entire cluster shutdown, especially when a single site with a majority of nodes fails, and the lack of communication paths results in resource management issues.
Innovation Solution
Each node in the cluster maintains information about the view of the cluster, allowing it to establish temporary sites and communicate with other nodes, enabling the election of new primary nodes and maintaining functionality even with a majority site failure, by using logical sites and proxy nodes for communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single quorum device is used for communication and vote exchange, then the voting mechanism is simplified, but the system reliability deteriorates because the quorum device becomes a single point of failure
Solution Approach 1:
The patent divides the cluster into multiple sites with distributed quorum devices at each site. Instead of relying on a single centralized quorum device, each site has its own quorum device that manages voting for nodes in that site. This segmentation eliminates the single point of failure while maintaining simplified local voting mechanisms.
2Ease of operation
If a majority voting scheme is used to select primary nodes, then the election process is straightforward, but the cluster fails completely when a majority of nodes in the primary site are lost
Solution Approach 1:
The patent segments the cluster into multiple sites, each with its own quorum device and primary node election independence. When nodes fail in one site, only that site is affected, not the entire cluster. Other sites continue to operate with their own primary nodes, maintaining cluster availability.
Solution Approach 2:
The patent implements beforehand cushioning by distributing quorum devices across multiple sites and maintaining backup primary nodes in each site. This preparation ensures that when node failures occur, the system has pre-established redundancy that prevents complete cluster failure.
3Productivity
If nodes are organized into sites with a primary node for resource management, then resource distribution is efficient, but the system cannot function when the primary site fails
Solution Approach 1:
The patent segments resource management into site-level quorum devices that independently manage resources within their sites. When one site fails, other sites continue to manage and distribute their resources independently, maintaining productivity while improving reliability.
Solution Approach 2:
The quorum device in each site serves multiple functions: it manages local primary node elections, coordinates resource distribution within the site, and maintains communication protocols. This multi-functionality allows the system to continue operating even when specific nodes or sites fail.
Data Source
AI summary
Resources may be assigned to nodes within a cluster of nodes and the nodes grouped to form sites. Each of the sites may elect a primary node to handle communications between the site and other sites within the cluster. The primary node of a site may be selected by a node identification number assigned to the node. When a node disconnects from a site, the node may form a temporary site and attempt to reestablish communications with the cluster. If communications are restored with the original site, then the node may rejoin the site. If communications with other sites are restored, then the node may form a new site. The dynamic cluster is tolerant to faults and failures within the nodes, because the nodes have information regarding the entire cluster. A site may remain operational with as few as one remaining node.