Order Book Management Device Dual-Core Architecture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional order book management devices experience high latency and overhead during update operations, which is exacerbated by the increasing command rate associated with 10G networks in financial exchanges.
Innovation Solution
The proposed solution involves an order book management device with a dual-core architecture, where the top-of-book management core handles the top part of the order book with low latency and the bottom-of-book management core manages the bottom part with a different data structure, optimizing processing performance and reducing latency variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If conventional order book management devices are used to handle update operations, then the system can process market data, but the latency increases significantly and overhead becomes important
Solution Approach 1:
The order book management device is divided into two independent cores: a top-of-book management core and a bottom-of-book management core. Each core handles specific portions of the order book independently, allowing parallel processing of update operations. This segmentation eliminates the bottleneck of sequential processing in conventional single-core devices, reducing latency while maintaining high command rate handling capability.
Solution Approach 2:
The invention introduces a dimensional change in data structure organization by separating the order book into top-of-book and bottom-of-book portions that can be managed in parallel dimensions. The top-of-book core manages the most recent and critical order book entries, while the bottom-of-book core manages historical entries, enabling simultaneous update operations across different dimensional layers of the order book.
2Productivity
If the command rate increases to handle 10G networks, then more market data can be processed, but latency variations increase and performance optimization becomes difficult
Solution Approach 1:
The system dynamically allocates processing resources between the top-of-book and bottom-of-book cores based on the incoming command rate and workload characteristics. The top-of-book core prioritizes low-latency processing for critical updates, while the bottom-of-book core handles less time-sensitive operations. This dynamic resource allocation allows the system to maintain optimized performance across varying command rates without excessive latency variation.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The invention provides an order book management device (10), implemented on an integrated circuit, for aggregating orders identified in input data streams into at least one order book, each order comprising a price and a quantity information, each order book being associated with a tradable object and comprising a sell side and a buy side, each order book side storing a list of limits, each limit being associated with a given price and a quantity, the quantity associated with a limit corresponding to the sum of the quantities of the received orders comprising the given price. Each side of the order book comprises : - a first data structure for maintaining a selected number P of limits corresponding to the P limits having the highest prices among the N limits maintained in the buy side of the book, or the lowest prices among the N limits maintained in the sell side of the order book, the orders in the first data structure being sorted by price depending on the order book side, and - a second data structure arranged in the form of a tree and comprising the N-P limits having the lowest prices among the N limits in the buy side of the book, or the highest prices among the N limits in the sell side of the book. The order book management device further comprises a first management core (2) for processing each received limit related command comprising a price information, updating the first data structure in response to the processing of the command and/or generating at least one update command to a second management core (3), the second management core (3) being configured to update the second data structure based on the update commands.