高頻交易系統需要大幅降低延遲以有效應對市場更新並實現利潤，進一步提升整體系統的延遲將潛在地增加高頻交易者的每日淨利。本研究旨在直接建立與交易所的連接，實施風險管理技術，並透過基於可程式化邏輯閘陣列(FPGA)的硬體加速實現更低的延遲。避開了第三方應用程式介面（API）的限制，交易者能夠減少延遲並充分利用市場機會。

研究者與合作廠商皓德盛科技有限公司提出一套雙向風險管理系統，運用現場可程式化邏輯閘陣列處理訂單監控、股票庫存控制與顧客資金控管，在必要時進行封包的替換。此外，我們開發了一個整合的回補程式與硬體協同執行，在每筆交易完成後即時進行更新。經過優化，該回補程式在一秒內能處理超過1000次交易執行。我們亦發展了一個多線程的交易程式，用於簡化下單過程以及訂單執行回報的迅速處理，提供硬體開發者系統驗證的環境。

本系統採用運行於 156.25 兆赫時脈頻率下的 10G 乙太網路實體收發器，以同步 64 位元乙太網傳送器和接收器之操作。研究結果顯示，所提出的系統在客戶至交易所方向的傳輸延遲為 550 奈秒，而交易所至客戶方向則為 582 奈秒，相較於傳統基於應用程式開發介面的交易方法，效率提升了一千倍之多。

High-frequency trading (HFT) systems require a significant reduction in latency to respond effectively to market updates and generate profits. Improving the overall system latency has the potential to increase daily net profits for high-frequency traders. This study aims to bypass third-party trading application programming interfaces (APIs), establish direct connections with exchanges, implement risk management techniques, and achieve lower latency through FPGA-based hardware acceleration. By bypassing the limitations of third-party application programming interfaces (APIs), traders can reduce latency and fully capitalize on market opportunities.

In collaboration with Vsense Fintech Inc., we have developed a proposed bidirectional risk management system that leverages FPGAs to manage order monitoring, position control, and fund manipulation. Additionally, an integrated compensation program has been designed to operate in conjunction with the hardware, executing updates upon the completion of each transaction. Through optimization, this compensation program demonstrates the capability to process over 1,000 executions within a single second. We have also created a multi-threaded trading program to facilitate order placement and the retrieval of order execution results. This program plays a crucial role in verifying the functionality of the hardware.

The implemented system utilizes a 10G Ethernet physical interface operating at a clock frequency of 156.25 MHz to synchronize the 64-bit Ethernet transceiver and receiver. The results indicate that the achieved latency of the proposed system is 550 ns in the client-to-exchange direction and 582 ns in the exchange-to-client direction. This outcome signifies an improvement in efficiency by a factor of one thousand compared to conventional API-based trading methodologies.

Abstract (Chinese) II
Acknowledgements (Chinese) IV
Abstract VI
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Main Contributions . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Organization . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Background Knowledge and Literature Survey 8
2.1 Financial Information eXchange (FIX) Protocol . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Message Format and Delimiter . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Session Flow . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.3 Message Categories . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Trading System . . . . . . . . . . . . . . . . . . . . . . . . 13

2.3 Risk Management . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Related Work . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Architecture Design 25
3.1 System Overview . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Risk Management . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Client to Exchange (CL2EX) Risk Management . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.2 Exchange to Client (EX2CL) Risk Management . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 Memory Management . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.1 Static Database . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.2 Dynamic Database . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.3 Dynamic Registers . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4 Host Access Interface . . . . . . . . . . . . . . . . . . . . . . . . 35
4 Software Implementation and Integration 37
4.1 Compensation Program . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Multi-threading Trading Program . . . . . . . . . . . . . . . . . . . . . . . . 44
4.3 Preload Program . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4 Web GUI . . . . . . . . . . . . . . . . . . . . . . . . 48
5 Evaluation Results 49
5.1 Hardware Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 49
5.1.1 Verification Methodology . . . . . . . . . . . . . . . . . . . . . . . . 49
5.1.2 Utilization of Hardware Resources . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.3 Performance Analysis . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 Software Optimization . . . . . . . . . . . . . . . . . . . . . . . . 58
5.2.1 Compensation Program . . . . . . . . . . . . . . . . . . . . . . . . 58
5.2.2 Preload Program . . . . . . . . . . . . . . . . . . . . . . . . 59
6 Conclusions and Future Works 61
6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 61
6.2 Future Works . . . . . . . . . . . . . . . . . . . . . . . . 62
Bibliography 64

[1] T. Hendershott, C. M. Jones, and A. J. Memkveld, “Does algorithmic
trading improve liquidity?” The Journal of Finance, vol. 66, no. 1, pp.
1–33, 2011. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.1111/
j.1540-6261.2010.01624.x
[2] T. Salkar, A. Shinde, N. Tamhankar, and N. Bhagat, “Algorithmic trading using
technical indicators,” in 2021 International Conference on Communication information and Computing Technology (ICCICT), 2021, pp. 1–6.
[3] Y.-C. Kao, H.-A. Chen, and H.-P. Ma, “An FPGA-based high-frequency trading
system for 10 gigabit ethernet with a latency of 433 ns,” in 2022 International
Symposium on VLSI Design, Automation and Test (VLSI-DAT), 2022, pp. 1–4.
[4] L.-H. Lin, Trends in High-Frequency Trading Regulation in the U.S. Securities
Market., 2016.
[5] C. Leber, B. Geib, and H. Litz, “High frequency trading acceleration using
FPGAs,” in 2011 21st International Conference on Field Programmable Logic and
Applications, 2011, pp. 317–322.
[6] P. Bressana, N. Zilberman, D. Vucinic, and R. Soulé, “Trading latency
for compute in the network,” in Proceedings of the Workshop on Network
Application Integration/CoDesign, ser. NAI ’20. New York, NY, USA:Association for Computing Machinery, 2020, p. 35–40. [Online]. Available:
https://doi.org/10.1145/3405672.3405807
[7] C. Păuna, “Capital and risk management for automated trading systems,” in Proceedings of the 17th International Conference on Informatics in Economy, 2018,
pp. 183–188.
[8] R. Martin, “Wall street’s quest to process data at the speed
of light,” 2007. [Online]. Available: https://rubinow.com/2007/04/21/
wall-streets-quest-to-process-data-at-the-speed-of-light/
[9] TradingView REST API Specification for Brokers (1.4.6), 2023. [Online].
Available: https://www.tradingview.com/rest-api-spec/
[10] A. Boutros, B. Grady, M. Abbas, and P. Chow, “Build fast, trade fast: FPGAbased high-frequency trading using high-level synthesis,” in 2017 International
Conference on ReConFigurable Computing and FPGAs (ReConFig), 2017, pp. 1–
6.
[11] Z. De-Hao, “A hardware-accelerated modular algorithmic trading system,”
Master’s thesis, 2022. [Online]. Available: https://hdl.handle.net/11296/xhr44u
[12] J. W. Lockwood, A. Gupte, N. Mehta, M. Blott, T. English, and K. Vissers, “A
low-latency library in FPGA hardware for high-frequency trading,” in 2012 IEEE
20th Annual Symposium on High-Performance Interconnects, 2012, pp. 9–16.
[13] D. Cui, “Financial credit risk warning based on big data analysis,” 2015. [Online].
Available: https://api.semanticscholar.org/CorpusID:204842293
[14] M. C. Miglionico and F. Parillo, “An application in bank credit risk management
system employing a bp neural network based on sfloat24 custom math library using
a low cost FPGA device,” in Advances in Computational Intelligence, S. Greco,B. Bouchon-Meunier, G. Coletti, M. Fedrizzi, B. Matarazzo, and R. R. Yager, Eds.
Berlin, Heidelberg: Springer Berlin Heidelberg, 2012, pp. 84–93.
[15] M. U. Torun, O. Yilmaz, and A. N. Akansu, “FPGA based eigenfiltering for realtime portfolio risk analysis,” in 2013 IEEE International Conference on Acoustics,
Speech and Signal Processing, 2013, pp. 8727–8731.
[16] Y.-C. Liao, “The FPGA implementation of a financial risk management system,”
2018. [Online]. Available: https://hdl.handle.net/11296/ekuy5y
[17] FAST Specification Version 1.1. FIXTrading Community, 2017.
[18] Taiwan Stock Exchange FIX 4.4 Message Specification Implementation Guide,
2022. [Online]. Available: https://dsp.tpex.org.tw/storage/regular_system/FIX4.
4%E9%9B%BB%E6%96%87%E8%A6%8F%E7%AF%84%E4%BD%9C%
E6%A5%AD%E6%89%8B%E5%86%8A.pdf
[19] T. J. Forum, Risk Management Practices and Regulatory Capital, 2001. [Online].
Available: https://www.iosco.org/library/pubdocs/pdf/IOSCOPD122.pdf
[20] A. H. Sameh, “On jacobi and jacobi-like algorithms for a parallel computer,” Mathematics of computation, vol. 25, no. 115, pp. 579–590, 1971.
[21] J. Jerome, G. Palmer, and R. Savani, “Market making with scaled beta policies,”
in Proceedings of the Third ACM International Conference on AI in Finance, ser.
ICAIF ’22. New York, NY, USA: Association for Computing Machinery, 2022,
p. 214–222. [Online]. Available: https://doi.org/10.1145/3533271.3561745
[22] Y.-F. Huang, “Low-latency warrants trading system of risk management with
speedy fast protocol decoder on NetFPGA,” Master’s thesis, 2019. [Online].
Available: https://hdl.handle.net/11296/u7w6yx
[23] R. Pagh and F. F. Rodler, “Cuckoo hashing,” Journal of Algorithms, vol. 51, no. 2,
pp. 122–144, 2004. [Online]. Available: https://www.sciencedirect.com/science/
article/pii/S0196677403001925
[24] G. Shi, M. Hu, H. Yin, W. Hu, and Z. Tang, “A shared virtual memory network
with fast remote direct memory access and message passing,” in 2004 IEEE International Conference on Cluster Computing (IEEE Cat. No.04EX935), 2004, pp.
495–.
[25] AMBA® 4 AXI4-Stream Protocol, 2010. [Online]. Available: https://developer.
arm.com/documentation/ihi0051/latest/