On July 28, Sinovel Technology A+CLUB and Amazon Cloud Technology held a salon event called "Summer Set Sail for the Cloud Technology Practice Closed-door Exchange Meeting" in Building 9 of Sinovel Technology, Shanghai, with nearly 30 quantitative institutions in Shanghai. Investment research and technical personnel conducted in-depth discussions and exchanges on how technologies such as software and hardware acceleration, cloud computing and HPC can be applied in quantitative investment.
Low-latency transaction optimization practice sharing
Huarui Technology Senior Solution Engineer Huang Minlong
With the development of the domestic quantitative industry, the application of low-latency trading technology in the entire securities trading chain continues to accelerate, and hardware acceleration technology is used in depth. In 2023, Sinovel will speed up and upgrade the entire trading chain such as trading desks, market quotes, offers, and APIs, integrating Optimization measures such as anti-cold state , multi-channel selection , and lightweight deployment have launched products such as ATP-IX Institutional Ultra-Speed (extremely fast trading), AMD-FPGA (extremely fast market conditions), and Sinovel Quote Machine , achieving a combination of software and hardware to ensure The entire transaction chain for institutional clients is stable and fast. At present, low-latency transaction optimization technology has entered the deep water area. In the future, Sinovel Technology will conduct multi-dimensional in-depth exploration and optimization of latency under the compliance framework, so stay tuned!
Using cloud computing to build a quantitative financial investment research environment
Amazon Cloud Technology Customer Solution Architect Zheng Mingming
As the founder and leader of global cloud computing, Amazon Cloud Technology has empowered millions of customers from all walks of life and can provide solutions to build a safe and controllable quantitative financial investment research environment, derivatives pricing and risk management calculation services on the cloud. , achieve low-cost accelerated verification of model factors, solve the IT resource and operation and maintenance problems faced by some domestic quantitative financial customers, and allow users to focus on strategy and business. On the cloud, users can use various computing, data storage and other quantitative technology stacks to build the required computing and data management platforms, and understand and manage investment research resources more clearly. It is worth mentioning that Amazon Cloud Technology has also designed training and inference instances (EC2 P5) for machine learning. Compared with the previous generation, the use of cloud computing has successfully helped customers shorten backtest task time and reduce costs. Compared with the original use of data Center training time is reduced by 6 times (from days to hours), saving 40% of costs. In the field of quantitative finance, seven of the world's top 10 quantitative fund management institutions are using Amazon cloud technology.
HPC construction experience sharing and cloud application exploration
Amazon Cloud Technology Senior Industry Solution Architect Liu Jianfeng
High-performance computing (HPC) is widely used in the field of quantitative finance and is suitable for pre-trade portfolio analysis, asset pricing, credit risk calculation and post-trade profit and loss calculation, market risk, asset and liability statements and other computing scenarios. Self-built HPC clusters There are common problems such as load unsaturation and capacity online restrictions. Amazon Cloud Technology cloud HPC can provide high-performance cluster access and task management with sub-millisecond latency , 400 GB/second throughput , millions of IOPS and 100,000-core concurrent access , and has an integrated HPC solution (SOCA) that can be developed twice . ). SOCA provides HPC basic environment and elastic resources, and has user interfaces, automation tools, customizable scripts and other tools to support resource configuration and permission isolation between teams, and customize self-service for unlimited resources in a reasonable way.