Distributed Storage Volume Migration Without Data Copying
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed storage systems, the inability to independently scale storage performance and capacity leads to resource inefficiencies and increased costs, as well as challenges in migrating data between servers while maintaining storage functionality without data copying.
Innovation Solution
A distributed storage system configuration where each server can directly access a shared drive box, allowing for independent scaling of storage performance and capacity, and enabling data migration between servers without copying data by managing metadata and using techniques like RAID configurations for data protection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If storage performance resource and storage capacity resource are stored together in the same server, then system integration is simplified, but storage performance and storage capacity cannot be independently scaled leading to resource inefficiency and increased system cost
Solution Approach 1:
The patent separates storage performance resources (CPU, cache memory) from storage capacity resources (drives) by placing them in different locations. Servers retain performance resources while drives are consolidated into shared drive boxes accessible by multiple servers, enabling independent scaling of each resource type without increasing overall system complexity
Solution Approach 2:
The drive box is designed as a universal storage resource that can be accessed and shared by multiple servers simultaneously. This multi-functional architecture allows any server to access any drive in the drive box, enabling flexible resource allocation and independent scaling of storage capacity without requiring dedicated storage for each server
2Productivity
If metadata related to volume is located only in the server having ownership of that volume to reduce communication frequency, then server autonomy is improved, but data migration between servers becomes difficult while maintaining storage functionality
Solution Approach 1:
The patent extracts the metadata from the source server and transfers it to the destination server during volume migration. This extraction approach allows the destination server to have complete knowledge of the migrated volume's data arrangement and configuration, enabling seamless continuation of storage operations without requiring ongoing communication with the source server
Solution Approach 2:
The system performs preliminary copying of metadata to the destination server before completing the data migration. This preliminary action ensures that when data migration is initiated, the destination server already possesses all necessary metadata information, allowing the migration to proceed without interrupting storage functionality or requiring complex coordination
3Adaptability or versatility
If data is migrated between servers for load distribution, then load balancing is improved, but network load and migration time increase due to data copying requirements
Solution Approach 1:
The patent copies only the metadata (data arrangement information) from the source server to the destination server, rather than copying the actual data. This selective copying approach allows the destination server to access and use the migrated volume immediately through direct drive access, eliminating lengthy data transfer operations and reducing migration time while still achieving load distribution
Solution Approach 2:
The migration process is segmented into two independent parts: metadata migration and data access. Metadata is migrated quickly through copying, while data access is maintained directly from the drive box without network transfer. This segmentation eliminates the need to copy actual data for migration purposes, significantly reducing network load and migration time
Data Source
AI summary
A volume to which a storage function is applied is migrated without copying data written to a volume as a migration object between computers while maintaining functionality of the storage function. A plurality of computers are accessibly connected to each of one or more physical storage devices. Each computer migrates ownership of the volume as the migration object to a computer as a destination. When the migration object volume from a first computer to a second computer is an owner volume to which a storage function is applied, the storage function requiring control data for I/O of data, the control data being metadata other than domain mapping data (indicating a relationship between a volume region and a storage region and being metadata for the owner volume), in place of or in addition to the domain mapping data.