Non-Volatile Memory Accessibility via Dynamic Data Relocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In shared computing environments where non-volatile media (NVM) is used as both memory and storage, the infrastructure topology impacts performance, leading to inefficiencies in data accessibility due to varying access patterns and network distances between processing nodes and NVM.
Innovation Solution
A method that monitors data requests to identify usage patterns of NVM as memory or storage, and modifies accessibility by optimizing network topology, such as reducing latency by relocating data or processes, replicating data, and balancing network loads based on these patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If data is stored in non-volatile media (NVM) in a shared computing environment, then storage capacity is improved, but data accessibility and latency are worsened due to varying network distances between processing nodes and NVM
Solution Approach 1:
The patent implements dynamic data migration that automatically relocates data between NVM regions based on real-time access patterns. When data is frequently accessed, it is migrated to NVM regions closer to the processing nodes, reducing access latency. This dynamic adjustment resolves the contradiction by making the storage system adaptive rather than static, allowing storage capacity to be maintained while access latency is optimized based on actual usage.
Solution Approach 2:
The patent introduces a fabric bridge as an intermediary device that monitors data access patterns and orchestrates data migration between different NVM regions. This intermediary analyzes access patterns and makes intelligent decisions about data placement, acting as a mediator between the storage subsystem and processing nodes to optimize the trade-off between storage capacity and access latency.
2Adaptability or versatility
If NVM is used for both memory and storage operations, then media utilization is improved, but performance is worsened due to competing access patterns and network congestion
Solution Approach 1:
The patent segments the NVM address space into multiple regions that can be independently managed and optimized. Different regions can be allocated for memory-like operations versus storage operations, allowing the system to handle both types of workloads simultaneously without them interfering with each other. This segmentation enables high-utilization media usage while maintaining performance by preventing access pattern conflicts.
Solution Approach 2:
The patent applies local quality optimization by tailoring the accessibility and performance characteristics of different NVM regions based on their specific usage patterns. Frequently accessed data regions receive optimized placement closer to processing nodes, while less frequently accessed regions can be placed in more distant NVM locations. This localized optimization allows the system to achieve high overall utilization while maintaining performance for critical operations.
3Reliability
If data is replicated across multiple NVM regions, then data accessibility is improved, but network bandwidth consumption is worsened
Solution Approach 1:
The patent implements feedback-based replication optimization where the system continuously monitors data access patterns and dynamically adjusts replication strategies. When data is frequently accessed from multiple nodes, replication is increased to improve accessibility. When access patterns become localized or less intensive, replication is reduced to conserve bandwidth. This feedback mechanism ensures that replication levels are optimized rather than static, balancing reliability improvements against bandwidth consumption.
Data Source
AI summary
Data requests for data stored in a non-volatile media may be monitored and used to identify if the media is being used as memory or storage. The accessibility of the data may be modified based on the identified usage model.