Mount Catalog for Rapid Volume Mount in Storage Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage systems face significant delays during partner node takeover due to the serial and time-consuming process of volume mounting, which is I/O bound and not scalable, especially in high-availability pair configurations, impacting the availability of data and performance in mission-critical applications.
Innovation Solution
Implementing a DBN-based mount catalog that stores storage device ID information and corresponding disk block numbers for each aggregate, allowing the takeover node to access the necessary blocks before determining the partner node's failure, thus avoiding the need for pre-fetching and disk reads, and maintaining the information in a reserved portion of the storage devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the existing volume mount process is used during HA pair takeover, then data integrity is maintained through serial reading of storage device blocks, but the volume mount time is excessive (5-10 seconds) causing delays in volume availability
Solution Approach 1:
The patent pre-loads volume mount information (superblock, block group descriptors, inode tables) into a mount catalog stored in memory before takeover occurs. This preliminary action eliminates the need for time-consuming serial disk reads during takeover, reducing volume mount time from 5-10 seconds to near-instantaneous while maintaining data integrity through the pre-cached information.
Solution Approach 2:
The patent creates a copy of the essential volume mount information from the storage devices and stores it in a mount catalog in memory. This copy contains all necessary data structures (superblock, block group descriptors, inode tables) needed for mounting volumes, allowing the takeover node to mount volumes without reading from the storage devices during the critical takeover window.
2Reliability
If the existing volume mount process is used during HA pair takeover, then proper failure detection is ensured by determining partner node status before mounting, but the takeover process is not scalable to N-way pairings and multiple volumes
Solution Approach 1:
The patent implements a universal mount catalog architecture that can serve any number of HA pair configurations (2-way, N-way) and any number of volumes. The mount catalog contains organized information that can be selectively accessed for any failed node, making the system scalable and adaptable to different HA pairing configurations without requiring fundamental process changes.
Solution Approach 2:
The patent segments the volume mount information into organized components (superblock, block group descriptors, inode tables) stored in the mount catalog. This segmentation allows efficient retrieval of only the necessary information for each specific volume during takeover, enabling scalable operation across multiple volumes and HA pair configurations without overwhelming the system with monolithic data structures.
3Measurement precision
If serial reading of random storage device blocks is performed during volume mount, then accurate volume information is obtained, but disk I/O bottlenecks occur making the process I/O bound and slow
Solution Approach 1:
The patent pre-loads all necessary volume information (superblock, block group descriptors, inode tables) into the mount catalog during normal operation before takeover is needed. This preliminary action ensures that during takeover, the system can immediately access pre-cached accurate volume information without performing slow serial disk reads, thereby eliminating I/O bottlenecks while maintaining information accuracy.
Solution Approach 2:
The patent introduces the mount catalog as an intermediary layer between the storage devices and the volume mounting process. This intermediary contains pre-cached copies of essential volume information, allowing the takeover node to obtain accurate volume information without directly reading from the storage devices during the critical takeover window, thus eliminating disk I/O bottlenecks.
Data Source
AI summary
Systems and methods which provide mount catalogs to facilitate rapid volume mount are shown. A mount catalog of embodiments may be provided for each aggregate containing volumes to be mounted by a takeover node of a storage system. The mount catalog may comprise a direct storage level, such as a DBN level, based mount catalog. Such mount catalogs may be maintained in a reserved portion of the storage devices containing a corresponding aggregate and volumes, wherein the storage device reserved portion is known to a takeover node. In operation according to embodiments, a HA pair takeover node uses a mount catalog to access the blocks used to mount volumes of a HA pair partner node prior to a final determination that the partner node is in fact a failed node and prior to onlining the aggregate containing the volumes.