Memory Data Migration Using Hot-Cold Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current memory data migration solutions are inefficient due to the need to migrate data in memory pages sequentially, which slows down the process and is further hindered by service processes that occupy memory resources, reducing overall efficiency, especially in high-service-pressure scenarios.
Innovation Solution
Classifying data into two categories: one for migration to another memory and the other for storage in a shared storage device, allowing for concurrent migration and storage, and adjusting priorities to ensure uninterrupted data transfer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is migrated in memory pages sequentially, then data integrity is maintained, but migration efficiency is reduced
Solution Approach 1:
The patent segments to-be-migrated data into two categories: first to-be-migrated data that must be migrated to another memory, and second to-be-migrated data that can be written to a storage device. This segmentation allows different migration paths for different data types, enabling parallel processing and improving overall migration efficiency while maintaining data integrity through classification-based handling.
2Adaptability or versatility
If service processes occupy memory resources, then system functionality is maintained, but migration speed is reduced
Solution Approach 1:
The patent extracts the migration process from the service process by implementing a dedicated memory migration apparatus that operates independently. The apparatus classifies and migrates data without being blocked by service processes occupying memory resources, thereby maintaining system functionality while achieving uninterrupted migration at improved speeds.
3Speed
If memory is used for direct operations, then data processing speed is improved, but memory damage risk increases
Solution Approach 1:
The patent introduces a storage device as an intermediary for the second category of to-be-migrated data. Instead of all data residing in volatile memory, frequently accessed data remains in memory for speed, while less critical data is written to persistent storage, reducing memory damage risk while maintaining data availability and processing speed for critical operations.
Data Source
AI summary
A memory data migration method, apparatus, and system are provided. During memory migration, data is classified into two parts based on a hot and cold degree of the data. Hot data is directly migrated, and cold data is written into a shared storage device shared by memories. When needing to be used in a destination-end memory, the cold data may be read from the shared storage device. This reduces an amount of data that needs to be migrated to the destination-end memory, thereby improving memory migration efficiency.


