FPGA Asset Management Device Parallel Hash Table Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current order management devices in financial data processing systems face performance limitations, particularly with high-throughput network connections, leading to increased latency and inability to process messages efficiently due to their reliance on general-purpose CPUs and software-based hash tables.
Innovation Solution
An asset management device and method utilizing a pipelined architecture based on FPGAs, which enables parallel processing of multiple messages through a hash table-like structure, optimizing clock frequency and memory usage to handle high-throughput data streams effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If general-purpose CPUs and software-based hash tables are used for order management, then ease of operation and development are improved, but processing speed and throughput are worsened
Solution Approach 1:
The patent replaces the software-based hash table implementation on general-purpose CPUs with a hardware-based hash table implemented in FPGA. This substitution of mechanical/software system with hardware system enables parallel processing of multiple hash table operations simultaneously, dramatically increasing processing speed while maintaining the ease of operation through configurable hardware design.
Solution Approach 2:
The patent segments the hash table into multiple independent banks that can be accessed in parallel. By dividing the single hash table structure into multiple concurrent access units, the system processes multiple messages simultaneously, resolving the contradiction between ease of operation and processing speed.
2Productivity
If the load is spread over several servers to handle high throughput, then message processing capacity is improved, but system latency increases
Solution Approach 1:
The patent merges multiple hash table operations into a single hardware-based hash table structure that supports concurrent access. By combining multiple processing capabilities into one unified hardware system, it achieves high message processing capacity without the latency penalty of distributed server architectures.
Solution Approach 2:
The hardware-based hash table enables continuous parallel processing of messages without the sequential bottlenecks of software implementations. Multiple messages are processed simultaneously in continuous streams, maintaining high productivity while minimizing latency through uninterrupted parallel operation.
3Speed
If FPGAs are used for data processing, then processing speed and parallel computation are improved, but device complexity and manufacturing cost increase
Solution Approach 1:
The patent implements a universal hardware hash table design that can be configured for different applications and data structures. The FPGA-based implementation provides multi-functional capability to handle various hash table operations (insertion, deletion, lookup) simultaneously, justifying the increased device complexity through enhanced processing speed and versatility.
Data Source
AI summary
An asset management method implemented on an integrated circuit uses a keys memory storing keys, each key being associated with an asset identifier, and a data memory storing asset information. The method comprises: receiving an input command for an asset comprising an asset identifier and asset information, computing addresses to Keys memory from the asset identifier, the computing addresses comprising calculating hashes from the asset identifier, finding or allocating an entry in keys memory for the asset, based on the computed set of addresses, depending on the input command, computing a data address to the data memory for the asset from the address and position in the keys memory at which an entry has been found or allocated for the asset; reading data in the data memory at the computed data address; and executing the input command based on the data read in the data memory at the data address.