Scale-out Storage Configuration Segmentation for High-Speed Failover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional scale-out-type storage systems face challenges in implementing high-speed failover due to the slow data access speed of nonvolatile data stores, which can lead to prolonged processing times and risks of delayed failover.
Innovation Solution
A storage system architecture where global configuration information is stored in a shared nonvolatile data store and local configuration information is held in-memory within subclusters, allowing I/O processing to occur without accessing the data store, enabling synchronization and maintaining consistency between global and local configuration information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If configuration information is stored in a nonvolatile data store, then data loss is prevented, but data access speed becomes extremely low
Solution Approach 1:
The patent divides configuration information into two types: global configuration information stored in the nonvolatile data store and local configuration information stored in memory. This segmentation allows each type to be stored in the most appropriate location, with local configuration information providing fast access during failover while global configuration information ensures data persistence.
Solution Approach 2:
The patent implements preliminary action by having the standby node acquire and store local configuration information from the active node before failure occurs. This pre-acquisition of configuration data in memory during normal operation enables immediate failover without needing to access the slow nonvolatile data store during the critical failover moment.
2Reliability
If configuration information is accessed during failover, then failover can be implemented, but processing time becomes long
Solution Approach 1:
The standby node performs preliminary acquisition of local configuration information from the active node during normal operation. When failover is needed, the standby node already has the necessary configuration data in its memory, eliminating the need for time-consuming access to the nonvolatile data store during the failover process itself.
Solution Approach 2:
The patent uses memory as an intermediary storage layer between the nonvolatile data store and the failover process. Local configuration information is cached in memory, which acts as a fast intermediary that can be accessed during failover without directly accessing the slow nonvolatile data store, thus reducing processing time.
3Stability of the object's composition
If all configuration information is stored in a shared data store, then consistency is maintained, but I/O processing performance decreases
Solution Approach 1:
The patent segments configuration information into global and local components. Global configuration information is stored in the shared nonvolatile data store for consistency across all nodes, while local configuration information is stored in each node's memory for fast access during I/O processing. This segmentation allows both consistency and high performance to coexist.
Solution Approach 2:
The patent applies local quality by storing configuration information locally in each node's memory rather than requiring all nodes to access a centralized data store. This local caching of configuration data enables high-speed I/O processing while the global configuration information in the shared store maintains overall system consistency.
Data Source
AI summary
Proposed is a scale-out-type storage system which implements high-availability, high-speed failover. In a scale-out-type storage system, two or more nodes each comprise a cluster controller, a node controller, a plurality of subcluster processes (subclusters and the like) which are processes which execute I/O processing in their own node, which form a subcluster between processes in their own node, and which are synchronized with work-type (active)/standby-type (passive) corresponding processes in the other nodes, and a nonvolatile data store (SODB). The configuration information of the storage system is held partitioned into global configuration information of the SODB and local configuration information and the like of the subclusters and the like, and thereupon the working-type subcluster is capable of executing I/O processing without accessing the SODB.