Storage Device with Integrated AI Computing for Data Transfer Speed
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current storage systems for AI computing face inefficiencies due to long data paths and slow data transfer speeds, as service data must be sent through networks multiple times between AI servers and storage devices, leading to low efficiency and slow data access during AI computing processes.
Innovation Solution
A distributed storage system is implemented with separate networks for AI parameters and service data, allowing direct high-speed interconnects within storage devices to reduce data transfer paths and leverage processing power from storage devices for AI computations, using PCIe, RDMA, and memory fabrics to accelerate data access and processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If service data is transmitted through network switches between storage devices and AI servers, then data can be accessed remotely, but data transfer speed decreases and transmission path length increases
Solution Approach 1:
The patent merges storage devices and AI servers into integrated storage computing devices, where storage units and computing units coexist within the same device. This integration eliminates the need for network transmission between separate storage devices and AI servers, directly shortening the data transmission path and increasing transfer speed by allowing the computing unit to access storage data through internal high-speed interfaces.
2Productivity
If separate storage clusters and AI clusters are used with network switching, then system scalability is improved, but data access efficiency deteriorates due to long transmission paths
Solution Approach 1:
The patent segments the storage computing device into independent storage units and computing units that can function separately or together. Each storage unit can serve multiple computing units within the same device, while the device itself can be part of a larger distributed system. This segmentation maintains scalability at the device level while improving data access efficiency through internal integration.
3Loss of time
If AI servers establish remote network connections to storage devices, then data can be retrieved from distributed storage, but transmission overhead increases and access time increases
Solution Approach 1:
The computing unit within the storage computing device acts as an intermediary between the storage unit and external AI servers. When data access is required, the computing unit can directly retrieve data from the storage unit through internal high-speed interfaces before processing or forwarding it, eliminating the need for AI servers to establish remote network connections to storage devices. This intermediary role significantly reduces both access time and network transmission overhead.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This application provides a storage device, a distributed storage system, and a data processing method, and belongs to the field of storage technologies. In this application, an AI apparatus is disposed inside a storage device, so that the storage device has an AI computing capability. In addition, the storage device further includes a processor and a hard disk, and therefore further has a service data storage capability. Therefore, convergence of storage and AI computing power is implemented. An AI parameter and service data are transmitted inside the storage device through a high-speed interconnect network without a need of being forwarded through an external network. Therefore, a path for transmitting the service data and the AI parameter is greatly shortened, and the service data can be loaded nearby, thereby accelerating loading.