Configurable Computational Storage Drives With FPGA Interface Switching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing compute-centric architectures struggle to scale with the increasing volume of data, and diverse storage devices complicate data access and analysis due to the need for multiple host controllers and interfaces, leading to redundant hardware and increased costs.
Innovation Solution
An FPGA-based system with a configurable drive interface that detects and adapts to various storage types, using a PCI bridge to convert communication protocols and support multiple drive interfaces, including SAS, SATA, and NVMe, eliminating the need for redundant host controllers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple host controllers are used to support diverse storage devices, then compatibility with various storage types is improved, but device complexity and hardware redundancy increase
Solution Approach 1:
The patent implements a universal host controller using FPGA technology that can dynamically configure and support multiple storage device interfaces (SAS, SATA, NVMe, USB) through a single device. The FPGA-based controller loads appropriate driver libraries based on detected storage device types, eliminating the need for multiple dedicated controllers while maintaining full compatibility across diverse storage protocols
Solution Approach 2:
The host controller employs dynamic configuration capabilities where the FPGA can reprogram itself at runtime to adapt to different storage device types. The system detects the connected storage device type and dynamically loads corresponding driver libraries, allowing the controller to transform its functionality based on real-time requirements rather than being statically configured for a single interface type
2Power
If compute-centric architecture is used, then data processing capability is improved, but scalability with increasing data volume deteriorates
Solution Approach 1:
The patent segments the computing architecture by implementing computational storage capabilities directly within storage devices through integrated FPGAs. This distributes processing power across multiple storage units rather than concentrating it in a central compute node, enabling the system to scale horizontally by adding more storage devices with embedded processing capabilities to handle increasing data volumes
Solution Approach 2:
The patent transitions from a traditional two-tier architecture (separate compute and storage) to a three-dimensional architecture where storage devices themselves possess computational capabilities. This adds a vertical dimension of computation within the storage layer, allowing data processing to occur at the edge of the storage system and enabling better scalability as data volumes grow
Data Source
AI summary
The system disclosed herein includes one or more storage drives, an FPGA based controller board, and a flash memory configured to store one or more FPGA executable binary libraries. The FPGA based controller board may include a drive logic detector configured to detect the type of the one or more of the storage drives and an FPGA executable libraries configuration module configured to select one or more of the FPGA executable binary libraries from the flash memory based on the type of the one or more of the storage drives to implement an FPGA based interface for communication between the one or more storage drives and a host.