Storage Device With Neural Processor For Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage devices face inefficiencies in data processing and transfer speeds, particularly with flash storage, due to the limitations of USB and PCIe protocols, and lack of integration with advanced hardware accelerators for enhanced functionality.
Innovation Solution
Incorporation of a hardware accelerator, such as a GPU, FPGA, or NPU, within the storage device to offload traditional processes, perform data indexing, encryption, and generate derivative assets using machine learning models, reducing the need for extensive data transfer and enhancing data manipulation capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data is stored and processed using traditional USB or PCIe protocols without hardware acceleration, then device compatibility and ease of connection are maintained, but data transfer speed and processing efficiency are limited
Solution Approach 1:
The patent combines storage functionality with hardware acceleration capabilities (GPU, FPGA, or NPU) into a single integrated device. The storage device includes both flash memory for data storage and a hardware accelerator for parallel processing, allowing data to be processed at the source without requiring separate processing hardware, thereby increasing data transfer speed and processing efficiency while maintaining protocol compatibility
Solution Approach 2:
The storage device is designed to perform multiple functions: it can store data using traditional USB or PCIe protocols while simultaneously providing hardware acceleration for parallel processing tasks. The device can adapt its functionality based on the connected system's needs, serving as both a storage device and a processing unit, which resolves the contradiction by enabling high-speed processing without requiring a completely new device architecture
2Productivity
If a hardware accelerator is integrated within the storage device to enable parallel processing and data manipulation, then processing efficiency and data access speed are improved, but device complexity and manufacturing difficulty increase
Solution Approach 1:
The patent introduces a controller as an intermediary component that manages communication between the flash memory and the hardware accelerator. The controller handles data routing, coordinate processing requests, and manage data flow between storage and processing units, thereby simplifying the integration process and making the device more manufacturable while still achieving high processing efficiency
Solution Approach 2:
The device is segmented into distinct functional modules: flash memory for storage, hardware accelerator for processing, and a controller for coordination. This modular segmentation allows each component to be optimized independently and facilitates manufacturing by enabling standardized interfaces and assembly processes, reducing overall device fabrication complexity while maintaining high productivity
3Adaptability or versatility
If extensive data is transferred between storage device and computer system for processing, then data manipulation capabilities are comprehensive, but data transfer time and system workload increase
Solution Approach 1:
The hardware accelerator processes data while it is still located in the storage device, performing parallel processing operations on stored data before it needs to be transferred to the computer system. This preliminary processing action reduces the amount of data that needs to be transferred and enables comprehensive data manipulation capabilities without incurring the time penalty of transferring extensive datasets
Data Source
AI summary
A storage device, comprises a storage controller, a flash memory and a neural processor. The neural processor is configured to retrieve a base asset from memory and generate a derivative asset using the base asset.