In-Storage Acceleration With Reconfigurable FPGA Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The processing of large-scale big data in data storage systems, particularly in NVMe and NVMe-oF devices, is costly and inefficient due to the need for heavy computing resources, large system memories, and high bandwidth networks, as data processing tasks are typically performed on remote host computers, leading to excessive energy and resource consumption.
Innovation Solution
Implementing an in-storage acceleration (ISA) platform in data storage devices like NVMe and NVMe-oF SSDs, which offloads data processing tasks to the storage devices using FPGA-based reconfigurable logic blocks and a configuration controller to execute data acceleration processes directly on the storage medium.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data processing tasks are performed on remote host computers, then data can be processed using existing computing resources, but computing resource consumption and energy usage increase significantly
Solution Approach 1:
The patent extracts data processing capabilities from the remote host computer and relocates them to the storage device itself. By implementing processing units and execution environments directly within the storage device, the system performs data processing locally at the storage location rather than requiring extraction of data to remote computers for processing, thereby reducing energy consumption while maintaining processing capability.
Solution Approach 2:
The patent introduces an intermediary processing environment within the storage device that acts as a mediator between data storage and remote host computers. This intermediary processing unit can execute data processing tasks locally, reducing the need for direct data transfer to remote computers and thereby lowering overall energy consumption while preserving processing functionality.
2Productivity
If data processing tasks are performed on remote host computers, then data can be processed using existing computing resources, but network bandwidth requirements increase significantly
Solution Approach 1:
The patent extracts data processing operations from the remote host and places them within the storage device. By performing processing locally at the storage location, the system minimizes the quantity of data that needs to be transmitted over the network, thereby reducing network bandwidth requirements while maintaining processing capability.
Solution Approach 2:
The patent implements preliminary data processing actions within the storage device before data needs to be accessed or transferred to remote host computers. By performing preprocessing, filtering, or transformation operations locally at the storage location, the system reduces the volume of data that requires network transmission, thereby lowering bandwidth requirements.
3Productivity
If data processing tasks are performed on remote host computers, then data can be processed using existing computing resources, but system cost increases significantly
Solution Approach 1:
The patent merges data storage and data processing functions into a single integrated system. By combining the processing units, execution environments, and storage media within the same device, the system eliminates the need for separate remote computing resources, thereby reducing overall system cost while maintaining processing capability.
Solution Approach 2:
The patent implements self-service processing capability within the storage device itself. The storage device can autonomously perform data processing tasks using its integrated processing units and execution environments, eliminating the need for expensive remote computing infrastructure and thereby reducing system cost while preserving processing functionality.
4Productivity
If data processing tasks are performed on remote host computers, then data can be processed using existing computing resources, but data movement between storage devices and host computer increases
Solution Approach 1:
The patent extracts data processing operations from the remote host and relocates them to the storage device. By performing processing locally at the storage location, the system minimizes the amount of data that needs to be moved between storage devices and host computers, thereby reducing data movement time while maintaining processing capability.
Solution Approach 2:
The patent performs preliminary data processing actions within the storage device before data needs to be transferred to or from the host computer. By completing processing operations locally in advance, the system reduces the volume of data that requires movement, thereby minimizing data movement time and improving overall efficiency.
Data Source
AI summary
A data storage device includes: a data storage medium; a processor comprising a plurality of processor cores; a plurality of application acceleration black-box (AABB) slots including reconfigurable logic blocks, interconnects, and memories; a host interface that receives a host command from a remote application running on a remote host computer, wherein the host command includes an image file including a register-transfer level (RTL) bitstream and a firmware driver; and a configuration controller that downloads the RTL bitstream to an AABB slot of the plurality of AABB slots and reconfigure the AABB slot, and load the firmware driver to a processor core of the processor. The processor core loaded with the firmware driver runs a data acceleration process of the remote application to access and process data stored in the data storage medium using the RTL bitstream downloaded in the AABB slot.