Financial markets are ever-changing, and trading opportunities are fleeting. Transaction speed has become a key factor in determining success or failure. Even a minor delay can miss the trading opportunity, affect the trading outcome, and lead to a loss of revenue. To meet the stringent requirements for data processing and transmission speed in the quantitative trading environment, Inspur Information has joined hands with Yuan Brain partners, Zhongke Yushu and Kuanrui Technology, to release the "Ultra-Low Latency Quantitative Trading Acceleration Solution" aimed at the securities industry.
The quantitative trading acceleration solution uses x86 architecture servers and Inspur Information's Cloud Lian operating system KOS as the underlying basic platform, equipped with Zhongke Yushu's SWIFT-2200 low-latency network card. After system-level optimization, it provides a stable, efficient, and ultra-low-latency basic support for Kuanrui Technology's rapid trading system. It has achieved full business live deployment from the securities institution's rapid counter system to the exchange's front-end system, reducing the overall network latency by 40%-50% and further accelerating the trading order.
Microsecond-level differences affect trading outcomes, and reducing network latency has become a necessity. The securities market is constantly changing, and trading opportunities are fleeting. Having a trading system with lower latency and faster response means that you can accurately grasp buying and selling opportunities and points. The faster the processing speed of the trading system, the more likely it is to gain an advantage in the rapidly changing market.
In the trading order link, network latency and trading system speed are particularly important, usually measured in microseconds (μs). Network latency is generally required to be below 1μs, and the internal penetration latency of the trading system is generally required to be around 4μs. When performance is good, it can reach 1-2μs. Pursuing ultra-low latency at the microsecond or even nanosecond level has become the key for securities companies to attract customers and expand the scale of brokerage business. Faced with this challenge, securities companies urgently need a latency acceleration solution that can meet real-time rapid response while also dealing with complex data processing and concurrent task requirements.
At present, latency acceleration solutions that rely solely on hardware or software have gradually shown their limitations. In latency-sensitive scenarios, pure hardware solutions are struggling to cope with dynamically changing system operation conditions. At the same time, as the manufacturing of chips and other hardware components gradually approaches the physical limit, further reduction in latency will become more difficult. Pure software solutions are constrained by traditional CPU design and algorithm complexity, making it difficult to fundamentally solve the latency problem. In addition, multi-tasking and concurrency are the norm in latency-sensitive scenarios, and pure hardware or pure software solutions often struggle to effectively handle resource competition and scheduling issues. Therefore, achieving ultra-low latency acceleration through software and hardware collaborative optimization technology has become an important development direction.
Inspur Information collaborates with Yuan Brain partners for collaborative optimization, reducing securities trading latency by 40-50%. To meet the high requirements of quantitative trading investors for trading speed, Kuanrui Technology has tailored a rapid trading system for users, with internal penetration latency as low as the microsecond level, leading the industry in full link latency. To achieve further acceleration of trading, Inspur Information, together with Yuan Brain partners Kuanrui Technology and Zhongke Yushu, has completed comprehensive optimization from underlying hardware to upper-layer applications. The underlying layer is the KOS operating system, integrating Zhongke Yushu's low-latency network card, and built-in BIOS parameters, network card interrupt aggregation parameters, interrupt affinity, and other best practice configurations, achieving fast transmission and processing of network data packets; the upper layer is combined with Kuanrui Technology's rapid trading system. After joint testing by the three parties, the network latency in the securities trading scenario is reduced by 40-50%.
Ultra-low latency quantitative trading acceleration solution:
■ Optimize BIOS parameter configuration to achieve the best performance and response time: The solution is based on KOS's fine-tuning of underlying hardware configurations. At the BIOS level, multi-dimensional mode adjustments and redundant configuration parameter removal measures are taken. Software and hardware work together to maintain a high-performance state, improving memory bandwidth and access efficiency, reducing latency and conflicts, and comprehensively enhancing multi-tasking and multi-threading processing performance in ultra-low latency scenarios.**Configure network card interrupt aggregation parameters to optimize applications in high network traffic scenarios:** In traditional settings, each network interface typically has an independent interrupt mechanism that triggers an interrupt response when a data packet arrives, increasing the CPU's workload. The solution optimizes the design to reduce redundant CPU overhead in handling interrupts and improve network throughput, thereby enhancing system performance and response speed in scenarios with ultra-low latency.
**Set interrupt affinity to improve performance and resource utilization efficiency:** Interrupt affinity refers to the technique of binding specific network card interrupts to specific processors or cores in multi-processor or multi-core systems. This helps optimize the system's interrupt handling performance and improve performance and resource utilization efficiency. The solution anchors business processes precisely to the CPU cores where network interrupts share L3 cache, reducing unnecessary CPU context switching during business execution and increasing cache utilization. This significantly reduces network transaction latency and enhances business processing performance.
Additionally, the solution integrates with Kuanrui Technology's low-latency, high-reliability, high-availability, one-stop quotation rapid trading system and employs the LAP latency analysis platform as a testing tool. It continuously collects various performance data in real-time, undergoing multiple optimizations and tests under high concurrency, high throughput, sparse single, pulse tests, and various extreme scenarios, demonstrating stable system performance.
Currently, this solution has been successfully applied in a well-known securities company, reducing the speed loss in network latency and other links, fully leveraging the ultra-low latency and low jitter characteristics of the Kuanrui rapid trading system, further reducing the speed of upstream order submission, and meeting the ultimate requirements of latency-sensitive users for trading speed. This successful case has set an example for the optimization and upgrading of core trading systems in the financial industry, highlighting the important role of hardware-software collaborative optimization technology in the construction of new generation financial infrastructure.
Looking to the future, the optimization and upgrading of core trading systems in the financial industry will shift from pilot practice to large-scale promotion. Inspur Information will continue to deepen cooperation with Kuanrui Technology and Zhongke Yushu, constantly promoting innovation and application of hardware-software collaborative technology, providing more reliable and efficient technical support for the financial industry, and contributing to the vigorous development of financial new infrastructure.