Persistent Memory Descriptor for Fast Data Restore
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional NVDIMM systems face inefficiencies in data backup and restore operations due to non-sequential data access from persistent memory, requiring software intervention and slowing down the restore process, as they lack guaranteed data alignment and sequential responses.
Innovation Solution
The implementation of descriptor fields, specifically Dynamic Descriptor Control Records (DDCR), which are embedded within data streams to track source information and ensure correct data alignment and location during restore operations, eliminating the need for sideband control signals and processor intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional NVDIMM systems perform data backup and restore operations without descriptor fields, then the system structure remains simpler, but software intervention is required and the restore process slows down
Solution Approach 1:
The data stream is segmented into discrete data units, each accompanied by a descriptor field containing address information. This segmentation allows the system to track and restore data to specific locations without requiring complex software intervention, thereby improving restore operation speed while maintaining manageable data structure complexity
Solution Approach 2:
Descriptor fields act as intermediary elements between the data units and the memory addresses. These descriptor fields carry address information that mediates the restore process, enabling the system to automatically place data in correct locations without software intervention, thus resolving the contradiction between speed and complexity
2Manufacturing precision
If persistent memory stores data without descriptor fields, then storage is more efficient, but data alignment and location accuracy are lost during restore operations
Solution Approach 1:
The descriptor field is merged with the data unit structure, combining address information and data into a single integrated unit. This merging ensures data alignment precision during restore operations while minimizing storage overhead, as the descriptor is stored alongside the data rather than as a separate structure
Solution Approach 2:
The system adds a dimensional layer by incorporating descriptor fields that contain address information alongside the data units. This dimensional addition provides the necessary alignment and location accuracy without significantly increasing the overall storage quantity, as the descriptors are compact and integrated with the data
3Reliability
If sequential data access is not guaranteed from persistent memory, then data access flexibility improves, but restore process reliability decreases due to non-sequential access
Solution Approach 1:
The descriptor fields provide feedback information containing address details for each data unit. This feedback mechanism ensures that even when data is accessed non-sequentially from persistent memory, the system can reliably track and restore data to the correct locations, maintaining restore process reliability while allowing flexible data access patterns
Data Source
AI summary
The present invention is directed to memory systems. More specifically, embodiments of the present invention provide a memory system with a volatile memory, a persistent memory, and a controller. In a save operation, the controller copies contents of the volatile memory to the persistent memory as data units with their corresponding descriptor fields, where the descriptor fields include address information. In a restore operation, the controller copies data units from the persistent memory to their corresponding locations based on addresses stored at descriptor fields. There are other embodiments as well.