Computational Storage Model Segmentation for Secure Space Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computational storage systems face inefficiencies in managing storage space and ensuring data security, particularly with large-capacity multimedia data, leading to performance deterioration and data loss due to inadequate storage management and external accidents.
Innovation Solution
A computational storage system that manages storage space efficiently by using a neural network model to generate inferred image data through dividing neural network models into sub-models and safely stores these models, allowing for recovery in case of loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a single neural network model is used for data processing, then processing capability is maintained, but storage space utilization deteriorates and data security is compromised
Solution Approach 1:
The patent divides a single neural network model into multiple sub-neural network models (first sub-neural network model, second sub-neural network model, etc.). Each sub-model is stored separately in the storage device, improving storage space utilization through efficient memory management. The segmentation allows the system to load only necessary sub-models into memory, reducing memory consumption while maintaining processing capability.
Solution Approach 2:
The patent pre-generates multiple sub-neural network models from the original neural network model and stores them in the storage device before they are needed. This preliminary action enables the system to quickly retrieve and use specific sub-models based on processing requirements, improving both storage efficiency and response time while providing redundancy for data security.
2Quantity of substance
If neural network models are divided into sub-models, then storage space utilization is improved, but model complexity increases
Solution Approach 1:
The patent introduces a computing device as an intermediary between the host device and the storage device. This computing device manages the complexity of handling multiple sub-neural network models by automatically selecting and loading appropriate sub-models based on processing requirements, coordinating between storage and computation, and simplifying the overall system architecture despite the segmented model structure.
3Quantity of substance
If data is stored in storage device, then storage capacity is sufficient, but data transfer speed becomes bottleneck
Solution Approach 1:
The patent extracts only the necessary sub-neural network models from the storage device into the memory of the computing device based on specific processing requirements. Instead of transferring entire neural network models or all available data, the system selectively loads only the required sub-models, reducing data transfer volume and improving transfer speed while maintaining sufficient storage capacity in the storage device.
Data Source
AI summary
Provided is a computational storage system including a storage device including a model storage configured to store a plurality of neural network models, and a computing device configured to generate inferred image data corresponding to original image data, based on the plurality of neural network models. The computing device is further configured to, in a process of generating the inferred image data, based on a number of a time a target neural network model from among the plurality of neural network models is used increasing by a first threshold value or more, generate at least one first sub-neural network model by dividing the target neural network model and stores the at least one first sub-neural network model in the model storage.