Lock State Management in Distributed Storage Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed storage systems, managing lock state information is challenging due to large volumes of data and limited memory resources in nodes, leading to bottlenecks and errors when processing requests, especially when nodes lack sufficient memory to store all lock state information.
Innovation Solution
A distributed storage system architecture where a meta-data volume (MDV) node serves as the authoritative source for lock state information, using novel lock state command messages to convey information between MDV and data volume (DV) nodes, allowing DV nodes to check local caches and query MDV for missing information, enabling efficient lock management and decision-making even with limited memory.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If lock state information is stored centrally in a repository, then lock management is simplified, but system performance degrades due to bottlenecks
Solution Approach 1:
The patent segments lock state information into two categories: hard lock state information (essential for correctness) and soft lock state information (nice-to-have). DV nodes store only hard lock state information locally, while soft lock state information remains centralized. This segmentation allows DV nodes to make lock decisions independently without querying the centralized repository, eliminating bottlenecks while maintaining proper lock management.
2Speed
If all lock state information is cached locally at DV nodes, then request processing speed increases, but memory resources are exhausted
Solution Approach 1:
The patent extracts only the essential hard lock state information from the complete lock state information and stores it locally at DV nodes. Soft lock state information is excluded from local caching. This extraction allows DV nodes to process requests quickly using locally cached hard lock information while conserving memory resources by not storing the full lock state information.
3Quantity of substance
If DV nodes have limited memory, then system resource utilization is optimized, but lock state management becomes error-prone
Solution Approach 1:
The patent applies local quality by making the DV node's local cache specialized for storing only hard lock state information, which is sufficient for making correct lock decisions. The MDV node maintains the complete lock state information as the authoritative source. This specialization ensures that DV nodes have reliable lock state information locally without requiring full memory capacity, preventing errors while optimizing memory utilization.
4Adaptability or versatility
If lock state information is distributed to all DV nodes, then system autonomy increases, but network traffic and complexity increase
Solution Approach 1:
The patent applies preliminary action by pre-distributing only hard lock state information to DV nodes before they need to make lock decisions. This preliminary distribution of essential information allows DV nodes to operate autonomously without requiring complex real-time communication with the MDV node for every request, reducing network traffic and system complexity while maintaining autonomy.
Data Source
AI summary
A system and method manages lock state information in a storage system. A meta-data volume node includes a lock state database which is a comprehensive source for lock state information about data containers in the system. A plurality of data volume nodes include local lock caches, which contain information about locks. Lock state messaging between the meta-data volume node and the data volume nodes is used to assign locks and to update local lock caches. The meta-data volume node is configured to determine whether input/output requests should be allowed or denied and to instruct the data volume nodes of this input/output operation result. Lock information is also sent to the data volume nodes for storage in local lock state caches to the extent the data volume nodes have the network capacity locally.