From June 2nd to 3rd, 2023, "2023 Quantitative Technology Carnival" was held at the Shanghai World Expo Center. This two-day quantitative technology event presented quantitative technological innovations in multiple directions and attracted a lot of attention.
On the second day of the carnival, several themed forums were held simultaneously.
The " Extreme Speed to the Future " theme forum hosted by Sinovel Technology invited experts from Shanyan Data, NVIDIA, Amazon Cloud Technology, Huatai Securities and other experts to discuss trading systems , quantitative technology , trading risk control and other industry hot topics.
Opening Speech by Speed to the Future
Miao Yong, General Manager of Sinovel Technology
Miao Yong, general manager of Sinovel Technology, said, " Technology changes transactions and embraces the future at an extremely fast pace ." The rapid development of technology is imperceptibly promoting the update and iteration of trading and risk control technology, injecting surging power into innovation in the trading field. As a technology-driven enterprise, Sinovel Technology deeply understands the importance of technological innovation, and will jointly promote progress and innovation in the trading field with an open mind and a cooperative spirit, and open up a broader future for the development of the industry.
Future Competitive Road of Extreme Trading
He Zhidong, Director of Sinovel Distributed Technology Laboratory
He Zhidong, director of Sinovel Distributed Technology Laboratory, shared his thoughts on the competition trend of ultra-fast transactions to the participants. In the field of trading, the balance between fast and stable is an unavoidable problem. He Zhidong pointed out that we are about to build on a stable foundation, and we need to pursue more extreme performance on the basis of stability . In recent years, the performance of industry transaction solutions has continued to improve, and the optimization of transaction links has also entered the stage of comprehensive deepening from comprehensive development.
In terms of improving the transaction speed of infrastructure, the trend reflected from last year to this year is the deeper use of infrastructure acceleration technology . For example, the accelerators of Intel’s fourth-generation Xeon servers and the latest overclocking servers are quickly used in trading scenarios, a layer of switches is widely used in market scenarios, and the deep acceleration technology of network cards is used in all aspects of trading markets, all of which reflect the industry’s enthusiasm for low-latency technology The use is getting more and more in-depth.
In addition, the trend of accelerating application hardening is also obvious. For example, there are more and more practical cases of FPGA trading systems, and the FPGA market is standard in the field of extremely fast trading today. Sinovel also released the FPGA extremely fast trading solution ATP-IF and the FPGA extremely fast market solution AMD-FPGA this year . These two solutions achieve extremely low latency in terms of performance, and at the same time achieve stability and speed through complementary hardware and software; and The API is fully compatible with the software solution, and users can seamlessly switch to use the FPGA solution.
"In 2021, we are chasing an improvement in the order of 50 tens of microseconds. In 2022, we are chasing an improvement in the order of 10 microseconds. In 2023, we will pursue an improvement in the order of 1 microsecond ."
In the highly competitive field of extremely fast transactions, Sinovel Technology emphasizes the importance of operation and maintenance . A lot of work has been done to ensure the operation and maintenance of the system to ensure the stability of the system.
Sinovel Technology is based on stability, focusing on the pursuit of ultimate performance , and constantly innovating in hardware technology , application hardening acceleration , operation and maintenance, etc., to cope with the future of extremely fast trading competition.
Real-time risk control technology innovation and exploration in the extremely fast trading environment
Huang Yuejun, Head of Pre-sales Solutions of Sinovel Technology
Huang Yuejun, head of Sinovel's pre-sales solutions, introduced Sinovel's latest product system and the innovation and exploration of real-time risk control technology in his speech at this forum.
Sinovel Technology is a company that provides overall solutions, providing supporting solutions for securities companies, public offerings, bank wealth management, insurance asset management and other industries. The product system covers core business systems such as fast trading, real-time risk control, fast market, real-time clearing, unified TA, and asset management platform.
Huang Yuejun mentioned that the bottom middleware of the company's products is completely independently developed, similar to the foundation of a house, providing stable and reliable bottom support for the entire application platform. The company has a distributed low-latency laboratory in the field of quantitative ultra-fast trading, and has a self-built research institute to provide compliance consulting and other related services. In terms of real-time risk control technology , the core service adopts a shared memory communication mechanism , which can conduct point-to-point message communication at a speed of nanoseconds, and uses RDMA acceleration technology to improve the overall performance of the risk control system.
Sinovel's real-time risk control platform covers the Shanghai-Shenzhen-North Stock Exchange , futures exchanges and the inter-bank market . Its business varieties include stocks , bonds , funds , futures , options , etc. It has been used in nearly 30 brokerage and self-operated business lines. put into production application.
In the current brokerage DMA revenue swap business scenario , the overall uplink full-link delay of Sinovel’s real-time risk control platform is less than 20 microseconds . Of course, the ultimate performance is always a direction that is constantly pursued, and Sinovel will continue to work hard to continuously empower the business scenarios of quantitative extremely fast transactions.
Roundtable Dialogue: Discussion on Speeding Up the Whole Link of Quantitative Investment
In the round table dialogue, it was hosted by He Zhidong , Director of Sinovel Distributed Technology Laboratory, Qiu Shanggao , CTO of Shanyan Data , Xi Yelei, Head of Huatai Securities Institutional Integration Service Technology Delivery Team, Shi Yongming , Director of Engineering and Solutions for NVIDIA Financial Industry , Amazon Cloud Technology Solutions Zheng Mingming, the solution architect, conducted in-depth discussions on "speeding up the whole chain of quantitative investment" and "technical challenges in the quantitative industry".
Quantitative investment involves multiple key links such as data mining, model building, strategy backtesting, strategy orders, and transaction execution. The construction of each link is inseparable from the underlying computing power facilities, the support provided by the basic software platform, and the upper-level trading applications. The solutions provided can ensure the efficient achievement of quantitative strategies by optimizing and synergizing all aspects of the transaction path. Regarding how different links play a role in the whole path of quantitative trading, Qiu Shanggao introduced Shanyan Data's technology that focuses on distributed processing of massive data. By using local SSD storage and deep learning models to apply to massive data, the algorithm effect is improved. . Shi Yongming mentioned that in the era of surging demand for computing power, NVIDIA provides stronger computing power support through continuously upgraded GPU cards, and provides development tools and SDKs that are easy to use, high performance, and scalable. Zheng Mingming introduced the cloud services provided by Amazon Cloud for financial customers, including resources and data support, as well as elastic resource pools and machine learning platforms on the cloud. Xi Yelei emphasized the advantages and technological leadership of Huatai Securities in ultra-fast trading, providing ultra-fast trading solutions covering the entire link of market and trading.
As for the case of efficiency improvement in the key links of quantitative investment, Qiu Shanggao shared the application case of Shanyan data in the field of deep learning, and solved the problem of irrational behavior by processing massive and complex data and deep learning models. Shi Yongming mentioned that NVIDIA, as a provider of accelerated computing solutions, provides full-stack accelerated computing solutions, including hardware and application-level acceleration. The host discussed the trend of deploying trading systems on the cloud. Zheng Mingming said that Amazon Cloud is cooperating with Nasdaq to explore cloud-native solutions. Xi Yelei pointed out that Huatai Securities' extremely fast solution has played a role in attracting business development and attracting the attention of customers.
Talking about the future competition and outlook of the quantitative industry. Xi Yelei pointed out that the future competition in the quantitative industry will focus on hardware technology, especially FPGA technology, which is widely used in the futures field. However, in the field of securities, the application of hardware technology is still in the stage of rapid development. He Zhidong added that in addition to pursuing higher transaction speeds, private equity clients also have higher requirements for service quality. One of the future focuses is to optimize according to the transaction models and characteristics of private equity customers to achieve better improvement. Shi Yongming looked forward to the future development direction of computing power, and believed that computing power will be promoted in two directions: the underlying hardware and the upper software. In terms of underlying hardware, the computing power of a single card will continue to increase, and it is necessary to solve the problems of high-speed communication and collaboration between cards, as well as the interconnection problems within the entire cluster. In terms of upper-layer software, it is necessary to improve ease of use, high performance and scalability, such as using CUDA-based NV programming language and acceleration library to accelerate tasks in different application scenarios. Zheng Mingming pointed out that the reason for the lack of application of cloud services in domestic quantitative investment is that the technologies and infrastructure used by different institutions and private equity clients are not uniform. However, as a young technology platform, the cloud has great potential in improving investment research efficiency and security. The cloud can improve the efficiency of backtesting tasks through full parallelization, accelerate the iteration speed of strategies, and provide high-performance computing and security guarantees to help reduce costs and protect customers' core code assets.
Overall, the quantitative trading industry has made remarkable progress under the pressure of technological development and competition. In the future, with the continuous upgrading of hardware technology and the wide application of cloud services, quantitative transactions will further improve transaction speed, computing power and decision-making efficiency. At the same time, cooperation and sharing will also become a trend, and the industry's cooperative ecology and resource sharing will also further promote the development and maturity of the industry.