Research Report on Technical Implementation Paths for Low-Latency Trading in China's A-Share Market in 2025
Posted Updated Quantitative Trading

Research Report on Technical Implementation Paths for Low-Latency Trading in China's A-Share Market in 2025

This report systematically reviews the key links in the low-latency trading stack for China’s A-share market in 2025 and beyond, from exchanges to brokers to the client side. Achieving extreme low latency requires coordinated optimization across trading units, ultra-fast broker counters, low-latency brokerage and market data services, networks, algorithms, and hardware. New regulatory rules also raise the bar for both technical implementation and compliance. By combining the latest technology and policy trends, the report evaluates the latency characteristics of different approaches and provides a reference for market participants.

This article was written by AI using publicly available information. Please exercise caution and independent judgment.

Overview of Low-Latency Trading in China’s A-Share Market: Current State and the 2025 Outlook

Definition and Measurement of Trading Latency in the A-Share Market

In financial-market terminology, trading latency refers specifically to the time difference between the generation of a trading instruction and the final execution confirmation returned by the exchange. In the A-share market, this covers the entire process from order initiation on the investor side, through broker systems, into the exchange matching engine, and then back from the exchange as an execution report. For high-frequency trading (HFT), strategies are characterized by extremely short holding periods and extremely high trading frequency, so latency sensitivity becomes extreme and is typically measured in microseconds (μs) or even nanoseconds (ns) 1. The core goal of “low-latency trading” is to minimize the time consumed at every link across the full trading chain 2.

It is important to note that the standard for “low latency” is itself evolving and depends on the use case. HFT pursues nanosecond-level speed, while for other forms of algorithmic trading even millisecond-level improvement may significantly improve trading performance. Looking toward 2025, the baseline for acceptable latency is expected to continue falling as technology advances.

At the same time, exchanges and regulators are paying increasing attention to the market impact of high-frequency and low-latency trading. In some jurisdictions, regulators have already introduced speed bumps or minimum resting-time requirements to balance market efficiency and stability 3. For example, the European Union once required high-frequency traders to maintain a delay of less than 500 milliseconds between quote submission and execution 3. Measuring end-to-end trading latency accurately and consistently is itself a challenge because the full order lifecycle spans multiple independent systems, including client systems, broker systems, and exchange systems. Standardized latency checkpoints are therefore crucial for meaningful technical comparisons.

The End-to-End Order Lifecycle and Key Latency Checkpoints

Order execution in the A-share market is a complex multi-stage process, and each link can introduce latency. Understanding the full lifecycle and its key latency points is the foundation of trading-speed optimization.

  1. Investor order generation: A trading strategy produces a decision based on market data and pre-defined logic, then converts it into concrete order parameters such as ticker, side, price, and quantity.
  2. Order transmission to the broker: The investor sends the order instruction to the selected broker through a trading terminal, such as a PC client or mobile app, an API, or direct market access (DMA).
  3. Broker-side processing:
    • Once the order enters the broker system, it first passes pre-trade risk checks such as fund and position validation.
    • It may then flow through an order management system (OMS) and execution management system (EMS) for processing and logging 4.
    • The broker’s internal network, server performance, and system architecture all affect latency at this stage.
  4. Order transmission to the exchange: Through the broker’s trading unit (PBU) and gateway, the order is sent to the relevant exchange, namely the Shanghai Stock Exchange, Shenzhen Stock Exchange, or Beijing Stock Exchange.
  5. Exchange matching engine processing:
    • Once inside the exchange system, the order first queues.
    • The matching engine processes instructions according to price priority and time priority 6.
    • The speed of the matching engine is a core component of exchange competitiveness.
  6. Execution confirmation: After the order is executed, the execution report returns from the exchange to the broker and is then pushed to the investor.

Take the Shanghai Stock Exchange (SSE) as an example. Investors must first open both a securities account and a capital account, select a broker branch as their agent, and complete designated-trading procedures. Orders may then be placed through counter entrustment, self-service terminals, telephone entrustment, or online trading. All transactions are matched automatically by the SSE’s computerized trading system, without manual intervention 8. SSE trading hours include the opening call auction on trading days from 9:15 to 9:25 and continuous trading from 9:30 to 11:30 and 13:00 to 15:00. During the opening call auction from 9:20 to 9:25, the exchange host does not accept order cancellations 6.

Trading hours at the Shenzhen Stock Exchange (SZSE) are similar, including the opening call auction from 9:15 to 9:25, continuous trading from 9:30 to 11:30 and 13:00 to 14:57, and the closing call auction from 14:57 to 15:00. During 9:20 to 9:25 and 14:57 to 15:00, the trading host does not accept cancellation instructions for auction trading 10.

The Beijing Stock Exchange (BSE) also specifies call-auction and continuous-trading periods as well as detailed rules for order submission and cancellation 11.

Every stage in the order lifecycle is a potential source of latency. Optimizing only one link, such as exchange connectivity, while ignoring others, such as broker-side efficiency, sharply reduces the overall benefit because latency accumulates throughout the chain. In addition, exchange-specific rules and time windows, such as auction periods and cancellation restrictions, directly determine what low-latency technology can actually do. Even with an extremely fast system, if an order cannot be submitted or modified within the time window allowed by the rules, the raw speed advantage becomes irrelevant.

The 2025 Outlook for Technology and Regulation

Looking ahead to 2025, low-latency trading in the A-share market is expected to continue evolving on both the technology and regulatory fronts.

Technology trends:

  • Deeper use of hardware acceleration: FPGA technology is expected to be used more broadly in trading systems and market-data processing to achieve ultra-low latency. Institutions such as Huatai Securities have already launched FPGA-based trading-counter systems 13. FPGA can deliver nanosecond-level processing in market-data decoding 14. In addition, networking technologies such as RDMA 15, along with optimized CPUs and low-latency NICs 16, will continue to be key for performance improvement.
  • The convergence of AI and low latency: As artificial intelligence is used increasingly in strategy generation 18, the execution layer must respond with correspondingly low latency.
  • Continuous architecture optimization: In-memory trading systems, kernel bypass, and similar techniques will continue to spread, reducing OS- and software-layer overhead 16.

Regulatory trends:

  • Tighter supervision of algorithmic and high-frequency trading: The China Securities Regulatory Commission (CSRC) and major exchanges are strengthening oversight of algorithmic and high-frequency trading 20. The Administrative Provisions for Program Trading in the Securities Market (Trial), effective October 8, 2024, together with subsequent exchange implementation rules, form the main regulatory framework 23 25.
  • Core regulatory elements: These include clear definitions of algorithmic trading and HFT, strict reporting obligations covering account information, funding source, strategy type, and server location, close monitoring of abnormal trading behavior, emphasis on fair market access, exploration of differentiated fee structures, and explicit requirements for technical systems.
  • A stronger role for exchanges: Exchanges are being given greater responsibilities in monitoring and management, such as managing trading-unit usage through flow-rate controls and traffic-based charging 32 and conducting real-time surveillance of algorithmic trading.

There is a dynamic balance between the technological race for speed and the regulatory goal of preserving fairness and stability. On one side, market participants will keep investing in R&D and applying frontier technologies such as FPGA, RDMA, and kernel bypass, continually driving latency lower and raising both the technical and capital barriers to competing at the top end. On the other side, by 2025 the regulatory framework around low-latency and HFT activity is likely to be more comprehensive and more detailed, bringing new compliance costs and operating constraints while aiming to protect the market’s overall health. Differentiated fees or restrictions aimed at specific forms of high-frequency behavior, for instance, may moderate pure speed competition to some extent.

Exchange Infrastructure: The Core of Speed

As market operators and the final venue of execution, exchanges provide the infrastructure on which low-latency trading is built. Trading units, trading seats, and exchange hosting services together form the bridge between investors and the market core.

Trading Units: The Channel That Reaches the Market Directly

A trading unit is the basic business unit granted by an exchange to a member or other recognized institution for participating in securities trading and receiving regulation and services. It is the technical carrier of trading permissions 10.

  • Access and types: Each member seat typically comes with one trading unit automatically. Additional trading units or special configurations must be applied for with the relevant exchange 6. At the SSE this is called a Participant Business Unit (PBU), which may be used by the member itself or, under certain conditions, leased to entities such as fund companies 36.
  • Flow rate and capacity management:
    • SSE: The exchange offers trading-unit flow-rate configuration services to prevent order congestion during high-concurrency periods. Each seat or newly purchased trading unit comes with some free flow-rate quota. One standard unit supports about 10 messages per second in practice, though actual throughput is based on real-time measured traffic, and one standard unit also means up to 20 unconfirmed orders can be in flight simultaneously. Flow rate beyond the free quota must be purchased 32.
    • SZSE: The exchange uses trading units to manage trading permissions 37. The latest program-trading rules focus on monitoring peak submission rates and daily order counts, especially for HFT. For example, activity above 300 order and cancellation messages per second, or above 20,000 per day, may be classified as high-frequency trading 28. The exchange may apply differentiated charges to such activity 28.
    • BSE: The Beijing exchange’s program-trading rules similarly mention monitoring the number and frequency of submissions and cancellations and potentially adjusting fees accordingly 29. Its HFT thresholds are similar to those of SZSE 29.
  • Technical services: Exchanges provide technical support for trading units, including setup, password management, and troubleshooting 36.

The flow rate of a trading unit directly determines the capacity ceiling for orders entering the exchange and is a key configurable factor affecting latency at the exchange gateway. When markets are volatile or trading volume surges, insufficient flow rate leads to queueing on the broker side or before the exchange gateway, increasing actual end-to-end latency even if every other link is fast. The SSE’s explicit flow-rate service was created to address such order congestion 32. For traders pursuing low latency, especially HFT firms, the exchange’s free flow-rate quota is typically insufficient. For instance, a regulatory HFT threshold of more than 300 messages per second is far above the roughly 10 messages per second of one free standard flow-rate unit. Buying additional paid capacity therefore becomes unavoidable and forms part of the cost structure of low-latency trading 28. However, as regulators focus more closely on submission speed, especially in HFT, simply buying maximum flow rate may be constrained by rules designed to prevent system abuse or market disruption. Traders must balance speed against compliance.

The following table summarizes key features of trading units across China’s main stock exchanges and looks ahead to 2025.

Table 1: Comparison of Trading Units Across China’s Main Stock Exchanges (2025 Outlook)

Feature Shanghai Stock Exchange (SSE) Shenzhen Stock Exchange (SZSE) Beijing Stock Exchange (BSE)
Unit name / type Participant Business Unit (PBU) Trading Unit Trading Unit
How obtained One per seat automatically; extra units may be requested or leased 34 One per seat automatically; additional units may be requested based on business needs 34 Members must apply to establish trading units 39
Flow-rate / capacity management Flow-rate configuration service; free quota of about 10 messages/second per standard unit; paid increments available; focused on order congestion 32 Program-trading rules monitor submission speed; HFT defined as >300 messages/second or >20,000/day; differentiated charges possible 28 Program-trading rules monitor frequency and volume; HFT threshold similar to SZSE; differentiated charges possible 29
Connectivity options Technical guidance provided by the exchange 36 Connects through gateways into the exchange trading system 34 Must comply with exchange technical specifications 39
Key latency implications and 2025 considerations Flow rate is a critical bottleneck; paid flow-rate cost matters; HFT speed limits may constrain use. In 2025, management is expected to become more refined and more integrated with program-trading oversight. HFT submission-rate limits are hard constraints; differentiated charges affect cost. Monitoring and traffic management are expected to become more mature and perhaps more dynamic by 2025. HFT rate limits and differentiated charges apply. As a newer market, BSE may gradually align more closely with mature-market trading-unit management while emphasizing service for innovative SMEs.

Trading Seats: Historical Legacy and Modern Meaning

A trading seat originally referred to a physical seat on the exchange floor, but the meaning has evolved. Ownership of a seat is the qualification for participating in exchange trading 48. In China’s A-share market, exchange members must hold at least one seat 37.

  • Seat types:
    • Physical seats and non-physical seats: A physical seat refers to a terminal on the trading floor where orders were entered manually. A non-physical seat refers to a broker connecting branch-office computer terminals directly to the exchange matching host through modern communications networks, enabling automatic order submission. Non-physical seats are far more efficient and materially reduce latency because they eliminate the manual reporting step 48. SZSE has fully adopted non-physical seats, and SSE has also shifted toward them 48.
    • A-share seats and B-share seats: Used respectively for A shares and B shares and related instruments 48.
    • Agency seats and proprietary seats: Agency seats are used when brokers execute trades for clients, while proprietary seats are used exclusively for the broker’s own securities business 48.
  • Relationship to trading units: Each seat automatically gives the member one trading-unit usage right 33. The trading unit is the concrete technical realization of trading access, while the seat is more closely tied to membership qualification and market-access rights.
  • Direct effect on latency: Because non-physical seats are directly connected to the exchange host and require no manual intervention, they inherently offer lower latency than physical seats 48. The distinction between agency and proprietary seats does not directly determine latency, but it may affect how a broker prioritizes resources internally for the order flow routed through associated trading units.

Although the trading unit is the direct technical interface, the non-physical seat is indispensable as the infrastructural basis of modern low-latency access. The move from physical to non-physical seats is a fundamental step in reducing latency across the system. The distinction between agency and proprietary seats is not itself a latency variable, but brokers may allocate internal resources more favorably to proprietary flow, indirectly affecting latency for different order types.

Exchange Co-Location and Proximity Services

Co-location means deploying trading servers within the exchange data center or extremely close to it, sharply reducing network latency to the matching engine by minimizing physical distance and network hops. It is one of the core deployment methods for HFT and other latency-sensitive strategies 1.

  • SSE, via SSETech:
    • Operates high-grade data centers such as the Jinqiao Data Center and the Securities Technology Building 51.
    • Provides rack rental, cross-connects, internal data-center interconnects, and the “Shanghai-Shenzhen high-speed channel” linking the SZSE southern center. It also offers local trading-network access and local market-data-network access so hosted users can connect directly to exchange trading and market-data systems 51.
    • SSETech has also built remote disaster-recovery systems for trading and market data at the SZSE southern center 51.
  • SZSE, via Shenzhen Securities Communication Co. (SSCC):
    • Operates the Binhai Data Center and Southern Data Center 54.
    • Provides both trading hosting, intended for users qualified to trade on SZSE and directly connect to the exchange trading system, and integrated hosting services 54.
    • SSCC also offers terrestrial leased lines and satellite connectivity to help clients connect to its data centers 55.
  • BSE:
    • The CSRC’s draft Administrative Provisions on Co-Location in the Securities and Futures Market states that exchanges must provide co-location services fairly, charge reasonably, and avoid discriminatory treatment 56.
    • Detailed BSE-hosting arrangements are less explicit in currently available materials than those of SSE and SZSE, though third-party data-center providers can provide general hosting services 57. HFT typically relies on co-location 50.
  • Network connectivity: This mainly includes leased-line services from telecom carriers such as China Mobile, China Unicom, and China Telecom 52, internal BGP Internet services inside the data center, and direct cross-connects 51.

Co-location is the cornerstone of ultra-low-latency trading in the A-share market because it provides the most direct and fastest physical path to the exchange matching engine. Differences in data-center choice, rack location, and cross-connect quality can all create small latency differences within the exchange ecosystem itself. Although regulators emphasize fair access 56, the technical complexity and cost of co-location objectively create a tiered access environment that favors participants with stronger capital and deeper technical resources.

The following table compares co-location services across China’s main stock exchanges.

Table 2: Comparison of Co-Location Services Across China’s Main Stock Exchanges (2025 Outlook)

Feature Shanghai Stock Exchange (SSE) via SSETech Shenzhen Stock Exchange (SZSE) via SSCC Beijing Stock Exchange (BSE)
Data-center operator / location SSETech; Jinqiao Data Center, Securities Technology Building, and others 51 Shenzhen Securities Communication Co.; Binhai Center and Southern Center 54 Expected to follow CSRC co-location provisions 56; may rely on third parties or gradually build its own or joint facilities
Core hosting services Rack rental, cross-connects, trading and market-data local access, Shanghai-Shenzhen high-speed channel 51 Trading hosting with direct exchange access, integrated hosting, rack rental 54 Services to be provided in line with co-location regulations, ensuring fair access 56
Network path to matching engine Ultra-low-latency internal networking and direct channels Ultra-low-latency internal networking and direct channels If co-location is provided, low-latency connectivity will be a design target
Expected latency reduction vs remote access Tens to hundreds of microseconds, or more Tens to hundreds of microseconds, or more Similar to SSE and SZSE if comparable hosting is offered
Key 2025 considerations Capacity expansion, faster internal interconnects, stronger disaster recovery, and tighter coordination with program-trading regulation Capacity expansion, stronger role of the Southern Center, upgraded network services, alignment with program-trading rules, and HFT support As the market develops, official hosting may be improved or expanded, with an emphasis on innovative SME needs and consistency with program-trading rules

Broker and Counter Infrastructure: Managing the Intermediary Layer

Between investors and exchanges, brokers are the key bridge. Their trading-counter systems, low-latency solutions, and direct market access services all have a direct and significant impact on final trading latency.

High-Speed Trading Counters: Competing for the Microsecond

High-speed counters are execution systems designed by brokers specifically for low-latency trading. Their goal is to minimize order-processing time inside the counter itself, typically measured in microseconds or even nanoseconds.

  • Overview of core technologies:
    • CTP (Comprehensive Transaction Platform) architecture: Originating in futures markets, its efficiency and stability have made it a reference architecture for some low-latency securities systems. Tests show that, under a hyper-converged architecture, one-way order-entry latency can reach 0.615 milliseconds, better than 0.9 milliseconds in a physical environment 62. Xiangcai Securities also reports microsecond-level latency 63.
    • UFT (Ultra Fast Trading): Represented by Hundsun’s UFT, this is one of the mainstream high-speed-counter solutions. Its latency has reportedly been reduced to 2 to 3 microseconds and it has been adopted by major brokers including Guotai Junan, CSC, GF Securities, and Huatai Securities 64. Huatai’s SMT-α channel, based on Huarui ATP technology, claims latency as low as 2 microseconds 64.
    • LDP (Low Delay Platform): For example, Huabao Securities has an LDP-based distributed low-latency architecture 66. China Merchants Securities also uses LDP technology, with core pass-through latency reportedly below 10 microseconds and more than 50% improvement in overall trading performance 67.
    • FPGA solutions: By hardwiring critical functions such as market-data decoding, risk control, and order generation into programmable hardware, FPGA can provide extremely low and deterministic latency. Huatai Securities has already launched an FPGA securities trading counter 13. Kingdom’s KFOS 3.5 system uses FPGA and reports overall pass-through latency of 1.7 microseconds and upstream counter pass-through latency of 2.1 microseconds 68. Shengli Technology also reports nanosecond-level market-data decoding engines and 2 to 3 microseconds end-to-end counter latency 14.
  • Major vendors and their latency performance:
    • Hundsun: UFT reports 2 to 3 microseconds 65; the newer high-speed PB system reports single-order response time of 4.35 to 5.95 milliseconds including 30 risk checks 69.
    • Kingdom: KFOS 3.5 reports 1.7 microseconds overall pass-through latency, 2.1 microseconds on the upstream counter path, and 2.4 milliseconds for full-path optimization 68.
    • Shengli Technology: 150 nanoseconds end-to-end for market data and 2 to 3 microseconds end-to-end for the trading counter 14.
    • AcceleCom: Futures-counter latency median of 1.5 to 2.7 microseconds 70, with entry into the broker market already underway 65.
    • Easysun Information: Core pass-through latency as low as 10 microseconds 67; its VBKR platform claims latency as low as 2 microseconds 71.
    • Xiangcai Securities: The Quant 360-MDS market-data system reports 3.5 microseconds internal latency and microsecond-level order execution 63. Measured full-path latency is 140 milliseconds for Shanghai and 2.3 milliseconds for Shenzhen 63.
  • Functional profile: High-speed counters usually focus on core execution functions such as order submission, cancellation, and report receipt, while supporting multiple A-share instruments including stocks, ETFs, bonds, and funds 63.

The wide range of high-speed counters now on the market, with their claimed microsecond-level internal latency, shows how intensely brokers are investing in the speed race. But internal counter latency is only one component of the full trading path. Actual end-to-end latency is much higher because it also includes network transmission, exchange processing, and other links. Xiangcai’s full-path numbers, Shanghai 140 milliseconds and Shenzhen 2.3 milliseconds, are much closer to real-world experience and also highlight differences between exchange access paths 63. FPGA is becoming a frontier technology in the high-speed-counter field, especially for tasks such as market-data decoding, where it is already moving toward nanosecond-level capability 13. That said, in the pursuit of ultimate speed, high-speed counters often keep functions relatively lean and focus on core instructions such as order entry and cancellation. More complex order types, broader risk management, and richer account services may still depend on slower but more feature-rich centralized trading systems 65, implying a trade-off between speed and functionality.

The following table compares mainstream high-speed-counter technologies in China’s A-share market.

Table 3: High-Speed Counter Technologies and Performance in China’s A-Share Market (2025 Outlook)

Vendor / Broker Product / System (examples) Core technology Claimed internal / core latency Markets / products supported Key low-latency features 2025 technology trend
Hundsun UFT UFT architecture 2-3 μs 65 Shanghai/Shenzhen A shares, funds, bonds, etc. In-memory database, optimized communications protocol Continued algorithm optimization, deeper FPGA integration, and broader low-latency PB services
Kingdom KFOS 3.5 FPGA, in-memory database, distributed architecture 1.7 μs overall, 2.1 μs upstream counter path 68 Shanghai/Shenzhen A shares, funds, bonds, etc. Hardware-accelerated risk control, parallel computing Broader FPGA coverage, further compression of full-path latency, smarter risk and execution logic
Shengli Technology High-speed counter / market-data system FPGA, low-latency network protocols 150 ns for market-data decoding, 2-3 μs counter E2E 14 Shanghai/Shenzhen A shares Hardware market-data decoding, full-path optimization Nanosecond technologies spreading to more links and more integrated market-data plus execution solutions
Huatai Securities FPGA counter, SMT-α FPGA, Huarui ATP SMT-α channel 2 μs 13 Stocks, funds, ETFs, bonds Full-path FPGA hardware and low-latency communications mode Parallel in-house and partner development, wider FPGA deployment, AI-plus-execution exploration
China Merchants Securities LDP high-speed trading system Low-latency distributed architecture (LDP) Core pass-through <10 μs 67 Shanghai/Shenzhen A shares Code optimization, compiler optimization, hardware tuning Continued LDP evolution, higher throughput and stability, broader business support
Easysun Information VBKR / low-latency trading platform In-memory technology, optimized algorithms VBKR as low as 2 μs 71, core pass-through 10 μs 67 Shanghai/Shenzhen A shares Microsecond-level processing engine Extending futures-market low-latency experience into securities and cross-market solutions
Xiangcai Securities Quant 360 (OES/MDS) CTP-like interface MDS market data 3.5 μs, microsecond-level order execution 63 Stocks, ETFs, bonds, funds, etc. on both exchanges Microsecond-level market-data service and order execution Ongoing end-to-end optimization, especially on SZSE access, finer flow-rate and subscription controls
General technology CTP-like counters CTP architecture Sub-millisecond to microsecond-level 62 Stocks, futures, etc. Mature and stable, widely accepted interfaces Further optimization under virtualized and cloud environments with more flexible deployment

Broker-Specific Low-Latency Solutions and DMA

To meet the extreme speed demands of institutions and quant investors, leading brokers are investing heavily in proprietary low-latency solutions and offering DMA services.

  • Broker-specific low-latency products:
    • CITIC Securities: Offers a memory-based “fast trading system” tailored for high-end clients 72. Its in-house CATS quantitative trading platform uses all-in-memory trading technology and claims microsecond-level latency, while also providing market data, data services, strategy development, algorithmic execution, risk control, and performance analysis 19. CITIC also offers co-location services 72.
    • Huatai Securities: Has launched an FPGA securities trading counter 13 and operates the SMT-α channel, based on Huarui ATP, with core processing latency as low as 2 microseconds 64. Huatai is also a user of mature solutions such as Hundsun UFT 65.
    • Guotai Junan Securities: Also uses systems such as Hundsun UFT 65 and has invested heavily in quantitative-trading services, with more than 50 counters including Guangce, Yida, Shengli, AcceleCom, and Hundsun NST. It has opened physical links to CFFEX, SSE, and SZSE and has deployed stock-option counters and cross-market depth data in exchange machine rooms for hosted clients 74.
    • China Merchants Securities: Uses the LDP high-speed trading system, with core pass-through latency below 10 microseconds 67. It offers direct market connection through DMA 76 and previously worked with Progress Software to launch a quant-trading platform based on Apama 77. Its institutional services also include fund custody and outsourcing 79.
    • Other brokers: Firms such as Galaxy Securities 81 and Shenwan Hongyuan Securities 47 also provide various trading services through their trading software and platforms, with some emphasizing fast execution.
  • Direct Market Access (DMA):
    • DMA allows client orders to be sent through the broker’s trading unit directly into the exchange matching engine, bypassing part of the broker’s normal internal order-processing stack, such as portions of a standard OMS or EMS, thereby reducing hops and cutting latency.
    • China Merchants Securities 76 and CLSA, in the Hong Kong context, 84 explicitly provide DMA services. Kingdom’s KFOS also supports DMA-like functionality 85.
    • DMA implementation usually relies on leased lines or VPN connectivity plus additional security controls, such as predefined connection IDs and restrictions on tradable accounts 86.

The heavy investment of leading brokers in low-latency infrastructure, including in-memory systems, FPGA, and deeply customized platforms such as CATS, reflects intense competition for high-end clients and has become an important differentiator in broker service capability. As a low-latency access method, DMA can materially reduce latency by cutting out internal broker hops. However, the exact implementation path and how “direct” the path really is can differ from broker to broker. Some mandatory risk-control steps may still pass through broker systems. Looking toward 2025, more brokers are likely to offer tiered low-latency services, with top-end solutions such as dedicated DMA channels and FPGA-accelerated paths available at higher cost for high-volume or highly latency-sensitive clients.

Key Supporting Technologies and Other Latency Drivers

Beyond the core infrastructure provided by exchanges and brokers, a set of supporting technologies and the client’s own system architecture are also crucial to total trading latency.

Market-Data Distribution and Processing

Timely, accurate, and comprehensive market data are the lifeblood of low-latency trading decisions. Every stage from market-data generation at the exchange to the moment it becomes actionable in the trading strategy affects reaction speed.

  • Exchange market-data sources:
    • SSE: Provides Level-1 data, Level-2 depth data with tick-by-tick trades and orders, ten-level order book, top-price queue depth, and FAST-formatted Level-1 feeds 87. Its Level-2 interface specifications include the SSE LDDS Auction Level-2 Interface Specification and the bond-market equivalent 88.
    • SZSE: Provides Level-1 and enhanced/depth Level-2 data 89. Its V5 market-data interface includes both C++ and Java APIs 89.
    • BSE: Provides Level-1 market data, and Level-2 may be available through authorized providers such as Shanghai Securities News Information services 53. Its market-data system primarily uses TCP for normal access and multicast for backup 100.
  • API protocols and latency:
    • Binary protocols: Compared with text protocols such as FIX, binary protocols usually provide lower latency because they are more compact and cheaper to parse. The SSE provides a binary market-data interface 101.
    • FIX (Financial Information eXchange): As a widely used global electronic-trading protocol, FIX is broadly adopted, but its text-based format can produce more redundancy and longer parsing time 102.
    • Proprietary APIs, such as CTP and UFT interfaces: These are usually binary and optimized for the vendor’s own system stack, with the goal of minimizing communication latency 63.
  • Hardware-accelerated market-data decoding: FPGA is widely used for ultra-low-latency market-data decoding and can compress this specific processing stage to the nanosecond level 14.

The speed and granularity of market data, such as Level-2 depth feeds versus Level-1, are critical for latency-sensitive strategies. Faster data means faster response, though deeper feeds such as tick-by-tick order streams also require stronger compute resources. Binary protocols and FPGA decoding represent the current low-latency frontier, though they often require more specialized engineering effort than standardized FIX. By 2025, exchanges may continue enhancing depth-data services and may provide paying clients with faster and more direct data interfaces, potentially through more advanced binary protocols or even direct memory access approaches.

The following table compares market-data delivery options in the A-share market from a latency perspective.

Table 4: Latency Impact of Market-Data Delivery Options in China’s A-Share Market (2025 Outlook)

Feature Shanghai Stock Exchange (SSE) Shenzhen Stock Exchange (SZSE) Beijing Stock Exchange (BSE)
Data types (examples) Level-1 (FAST, Binary), Level-2 (LDDS) 88 Level-1, Level-2 (V5 interface, C++ / Java API) 89 Level-1 (TCP / multicast), Level-2 (possibly through authorized providers) 88
API protocols (examples) Binary, FIX via FAST, proprietary LDDS Proprietary V5 API, Binary TCP/IP, proprietary
Typical latency (exchange to client receipt) Depends on access mode, such as co-location or leased line, and on client-side processing capacity. In co-location, the target is microseconds. Similar to SSE; co-location and leased-line access are key. Similar to SSE and SZSE; physical proximity is the major factor.
Bandwidth demand Higher for Level-2 and tick-by-tick data Higher for Level-2 and tick-by-tick data Higher for Level-2 data
Low-latency hardware/software FPGA decoder cards 14, low-latency NICs, optimized parsing libraries FPGA decoder cards, low-latency NICs, optimized parsing libraries FPGA decoder cards, low-latency NICs, optimized parsing libraries
Pros and cons for HFT Pros: rich depth data and multiple interface choices. Cons: obtaining the very lowest-latency feed is expensive. Pros: convenient V5 interface and depth data. Cons: high bandwidth and processing demands. Pros: more room for innovation in an earlier-stage market. Cons: Level-2 coverage and interface maturity may lag SSE and SZSE.
2025 trend Faster push frequency, finer-grained depth content, and possibly RDMA-based market-data interfaces Unified and optimized multi-market interfaces, improved API usability and performance, and exploration of compression techniques to reduce bandwidth Continued improvement in Level-2 coverage and interface standards, stronger market-data-system stability and processing capacity, and encouragement for third-party providers

Network Infrastructure Optimization

The network is the nervous system connecting trading participants, brokers, and exchanges, and its performance directly determines transmission latency.

  • Leased lines: Compared with the public Internet, dedicated leased lines provide private, high-bandwidth, low, and stable latency connections and are the preferred option for reaching exchange hosting facilities or broker DMA access points 51.
  • Low-latency network equipment: Switches, routers, and similar hardware designed specifically for low-latency operation minimize internal packet-processing delay.
  • Optimized NICs:
    • SR-IOV (Single Root I/O Virtualization): In virtualized environments, SR-IOV lets virtual machines share a physical NIC directly, bypassing the host’s virtual-switch layer and reducing virtualization-induced network latency 16.
    • Kernel bypass: Technologies such as DPDK and custom drivers allow user-space applications to talk directly to NIC hardware, bypassing the operating-system network stack and substantially reducing OS-induced delay 16.
  • Protocol choice and optimization:
    • TCP: A connection-oriented reliable transport protocol, but handshakes, acknowledgements, and congestion-control behavior all add overhead and latency. Poor congestion handling may also produce bufferbloat 108. Some SSE market-data and trading paths use TCP 51, and BSE market data primarily uses TCP 100.
    • UDP: A connectionless protocol with lower overhead and faster performance but lower reliability, because delivery and ordering are not guaranteed. It is commonly used for market-data distribution, where speed matters most and a small amount of packet loss can be handled at the application layer 51.
  • Physical proximity: Minimizing physical distance is the most fundamental way to reduce network latency, which is exactly why co-location matters so much 1.

Optimizing every layer of the network stack, from leased lines and switching hardware at the physical layer, to tuned NICs at the link layer, to protocol choice and kernel bypass at the transport layer, is crucial for eliminating critical microseconds. The choice between TCP and UDP requires balancing reliability and speed, especially for market data. For order-entry paths, TCP’s reliability is usually preferred, but it is often wrapped in lean binary protocols to reduce its latency cost. By 2025, kernel bypass and SR-IOV are likely to be more widely used across trading infrastructure, and more mature network telemetry tools may emerge for real-time latency monitoring and control.

Client-Side System Architecture

The client’s own system design, hardware-software stack, and algorithmic efficiency are the final internal determinants of trading latency.

  • Trading algorithms and OMS / EMS design:
    • Efficient strategy logic: The computational complexity of the trading algorithm itself should be kept as low as possible, avoiding unnecessary processing.
    • OMS / EMS performance: The internal processing speed of the order management and execution management systems, from receiving a signal and generating an order to running risk checks and tracking state, directly affects total latency.
  • Hardware and software stack tuning:
    • CPU: Use high-performance CPUs and pin critical trading threads to specific cores using CPU affinity to improve cache hit rates and reduce context-switching overhead 17. Disable CPU power-saving states such as C-states and P-states so the processor runs at a stable high frequency, avoiding jitter from frequency changes 17.
    • Memory: Use fast RAM, optimize memory-access patterns, and use huge pages to reduce TLB misses 17. CITIC Securities’ CATS platform explicitly mentions an all-in-memory trading architecture 19.
    • Operating system: Use real-time patches and kernel tuning to minimize scheduling delay and interrupt jitter. Linux is a common and deeply optimized choice 110.
    • Programming language and software implementation: Compiled languages such as C++, or Java after ahead-of-time compilation, are generally preferable to interpreted languages. Efficient data structures and algorithms are essential.
  • FPGA on the client side: In addition to market-data decoding, FPGA can also be used client-side for ultra-latency-sensitive trading logic or pre-trade risk control modules 14.

Client-side optimization is just as important as network and broker-side optimization. Inefficient algorithms, poorly tuned hardware or operating systems, and slow internal message-passing mechanisms can erase speed gains achieved elsewhere. The trend toward all-in-memory trading systems 19 and client-side hardware acceleration such as FPGA 14 shows that the battle for nanoseconds has already moved deep into the trader’s own infrastructure. By 2025, client systems built for ultra-low latency are likely to become even more complex and require highly specialized engineering talent. At the same time, wider use of AI in signal generation 18 will impose even tighter latency demands on the path from strategy signal to order execution.

The Regulatory Framework for Low-Latency and Program Trading in 2025

China’s A-share market is entering a more structured and explicit regulatory era for program trading and low-latency trading. The CSRC and major exchanges have introduced a series of regulations and implementation rules intended to regulate market behavior, contain risk, and safeguard fairness and stability.

  • Core regulations and guidance:
    • In 2024, the CSRC issued the Administrative Provisions for Program Trading in the Securities Market (Trial), effective October 8, 2024, establishing the overall framework for program-trading oversight 25.
    • The SSE, SZSE, and BSE then issued, or are expected to issue, their own implementation rules for program-trading management to operationalize the higher-level framework 22.
  • Key regulatory provisions:
    • Definitions of program trading and high-frequency trading: Program trading is broadly defined as trading in which computer programs automatically generate or submit instructions. HFT, as a subset of program trading, is characterized by extremely high submission and cancellation speed and frequency. For example, the three main exchanges classify activity such as more than 300 order and cancellation messages per second on a single account, or more than 20,000 such messages per day, as high-frequency trading 23.
    • Reporting regime: The principle is “report first, trade later.” Program-trading investors must report basic account information, funding details including source and leverage, the type and main content of the strategy, order-execution method, peak submission rate, maximum daily order count, software information, and similar data to the broker. After the broker checks and confirms the submission, it reports to the exchange, and only then may the client trade programmatically 25. HFT participants must also report server location, system test reports, fault-response plans, and similar information 23.
    • Fair access and market stability: When exchanges provide hosting services, they must follow principles of safety, fairness, and reasonableness, distributing resources fairly and avoiding discriminatory treatment 56. Regulators seek to prevent market manipulation and ensure exchange-system safety and orderly trading 20. Brokers must not provide special convenience to program-trading investors when granting trading-unit services 25.
    • Monitoring and risk control: Exchanges conduct real-time monitoring of program trading and focus on abnormal activity that may affect system safety or trading order, such as abnormal peak submission rates, frequent rapid cancellations, repeated pushing up or pressing down of stock prices, or large transactions within a short period 22. Brokers must manage client behavior, and institutional investors must establish sound compliance and risk-control systems 25.
    • Differentiated charging: Exchanges may impose differentiated charges on HFT, for example by raising traffic-related or cancellation fees based on message count and frequency or increasing market-data charges 23.
    • Technical-system requirements: Systems used for program trading must meet exchange requirements and provide functions such as verification of funds and securities, permission control, threshold management, abnormal-behavior monitoring, error handling, and emergency procedures, while ensuring stable and secure operation. Sufficient testing is required before use, and test records must be reported 25.

From 2024 into 2025, regulation of program trading and HFT in the A-share market is clearly becoming more structured and more stringent. On one hand, the new rules unify market expectations and provide clearer guidance for related trading activity. On the other hand, they raise compliance costs and potentially direct trading costs, for example through differentiated fees. Fairness is a recurring theme. Regulators are trying to balance the technical advantage sought by HFT and low-latency participants against the need for an orderly market shared by all participants, which may lead to tighter controls on strategies deemed to abuse technological advantage. The detailed reporting regime will also give regulators unprecedented transparency into program-trading activity, improving supervision and laying a foundation for future data-driven policy adjustments.

The following table summarizes key rules affecting low-latency and program trading in the A-share market.

Table 5: Key Regulatory Rules for Low-Latency and Program Trading in China’s A-Share Market (Effective in 2024 and 2025)

Regulator Rule / guidance (examples) Key provisions (examples) Impact on low-latency strategies Expected 2025 focus
China Securities Regulatory Commission (CSRC) Administrative Provisions for Program Trading in the Securities Market (Trial) 25 Defines program trading, investor categories, reporting obligations, broker responsibilities, abnormal-trading monitoring principles, HFT supervisory direction, system requirements, and cross-border supervision principles Establishes the legal and supervisory foundation of program trading while increasing compliance requirements and potential cost Ensure full implementation, strengthen broker enforcement of client-management duties, push market participants to build internal control systems, and continue monitoring HFT’s market impact
Shanghai Stock Exchange (SSE) SSE Implementation Rules for Program Trading Management (Trial) 23 HFT criteria such as >300 messages/second or >20,000/day, extra reporting on server location, test reports and emergency plans, focused monitoring of abnormal behavior such as unusual peak rate and frequent rapid cancellations, and authority for differentiated charging Imposes more concrete compliance and cost requirements on HFT and constrains extreme behavior Refine abnormal-behavior monitoring metrics, implement differentiated charging, and strengthen checks on system stability and emergency capability
Shenzhen Stock Exchange (SZSE) SZSE Implementation Rules for Program Trading Management 26 Similar HFT thresholds and additional reporting requirements, clear abnormal-behavior monitoring, authorization for differentiated charging, and emphasis on member responsibility for client program trading Similar impact to SSE, bringing major markets under a more unified supervisory approach Tighten checks on consistency between strategies and reported information, improve monitoring precision through technology, and watch the effect of program trading on liquidity and volatility
Beijing Stock Exchange (BSE) BSE Implementation Rules for Program Trading Management (Trial) 29 HFT thresholds and extra reporting largely aligned with SSE and SZSE, clear abnormal-trading criteria, and authority for differentiated charging Provides explicit rules for BSE program trading and harmonizes oversight with other boards Support financing and growth of innovative SMEs while ensuring orderly trading and gradually refining a BSE-specific supervisory mechanism

Comparative Analysis and Strategic Recommendations for Achieving Minimum Latency in 2025

Achieving the lowest possible trading latency in China’s A-share market in 2025 requires a comprehensive view of technical paths, infrastructure choices, and the evolving regulatory environment.

  • Latency comparison across technical paths:
    • Path 1: Exchange co-location + DMA + FPGA-accelerated counter or client system. This is theoretically the lowest-latency top-tier configuration. By placing the system inside the exchange data center, physical network distance is minimized. A DMA path bypasses much of the broker’s traditional counter processing. FPGA provides hardware acceleration for market-data decoding, order generation, and risk control. This path is also the most expensive and technically demanding.
    • Path 2: Exchange co-location + broker-provided ultra-fast counter. This remains a very low-latency solution, but its performance depends on the broker’s counter technology and the network quality inside the co-location facility. Latency depends on whether the counter uses FPGA, how far the broker has optimized it for in-memory processing, and what the internal network path from co-location rack to core counter looks like.
    • Path 3: Remote DMA, where available, plus an optimized wide-area network. Latency is higher than in a co-location setup because the order still travels over a WAN, but it is usually lower than standard remote counter access because broker-side processing is reduced.
    • Path 4: Remote access through a standard broker platform. This is the highest-latency path because the order traverses the public network and the full centralized broker trading system.
  • Recommendations on infrastructure choices:
    • HFT and market-making strategies: Exchange co-location is essentially mandatory. Priority should be given to brokers with clean DMA access, along with in-house or third-party FPGA-based client systems and market-data systems.
    • Other latency-sensitive algorithmic trading: Co-location remains preferable, and pairing it with a high-quality broker counter or stable DMA path is usually sufficient.
    • Best execution for institutional orders: The absolute minimum latency may not be the main objective, but these workflows can still benefit from smart order routing, reliable DMA, or efficient broker execution algorithms.
  • Key factors in selecting vendors and brokers:
    • Verifiable latency data: Distinguish among counter-internal latency, broker-level end-to-end latency, and full-path exchange latency. Require clear measurement methods and test conditions.
    • Technical sophistication: Whether the solution uses FPGA, kernel bypass, in-memory architecture, and similar frontier techniques.
    • API quality and protocol efficiency: Whether interfaces are stable and efficient and whether they use lean binary protocols rather than standard FIX.
    • Co-location resources and connection quality: The broker’s presence inside exchange data centers, available rack resources, and network quality to exchange core systems.
    • Market-data quality and speed: Depth-data availability, speed, and API quality.
    • Compliance and transparency: Whether the service can satisfy the latest program-trading requirements and whether terms are clear.
    • Technical support and domain expertise: The responsiveness and professionalism of the vendor or broker technical team.
    • Cost: Access fees, flow-rate fees, hosting costs, and software-license fees all matter.
  • Forward-looking strategy for 2025:
    • Modular and scalable architecture: Invest in system designs that can be upgraded and scaled flexibly as technology and business change.
    • Use openness where appropriate, focus on the core where it matters: Open standards may be suitable for general-purpose components, but for decisive low-latency links, proprietary high-efficiency technology may be worth adopting.
    • Track emerging hardware: Continue monitoring the use of FPGA, ASIC, and specialized networking processors such as DPU or IPU in trading 15.
    • Build in compliance from the start: Reporting and risk-control requirements should be designed into the system and workflow, not bolted on afterward.
    • Integrate AI with execution: As AI becomes more central in trading strategies, execution systems must be able to react to AI-generated signals with low latency 18.

Reaching the lowest latency in 2025 will not come from a single breakthrough or one-off infrastructure upgrade. Real leadership comes from a carefully engineered systems approach spanning physical deployment, network connectivity, trading interfaces, counter technology, client algorithms, and hardware acceleration all at once. This is an interlocking chain, and any weak link can become the performance bottleneck.

At the same time, the marginal cost and technical difficulty of each further reduction in latency rise sharply as latency falls. Moving from milliseconds to microseconds, and then from microseconds to nanoseconds, typically requires exponentially greater effort in technology, hardware, and talent. Institutions therefore need to set latency targets based on how sensitive their own strategies truly are to speed and whether the investment is justified.

Finally, the program-trading framework that took effect in 2024 and will keep evolving through 2025 must be treated as a core background condition for all technical decisions 23. Pursuing extreme speed must go hand in hand with compliant behavior, stable systems, and respect for market fairness and order. The low-latency trading system of the future must be not only fast, but also intelligent and responsible. This means leaving interfaces and functions in the system for regulatory reporting and for adapting flexibly to future changes in rules or market structure.

Conclusion and Recommendations

Pursuing the lowest possible trading latency in China’s A-share market is a systems-engineering problem that requires deep understanding and careful optimization across exchange infrastructure, broker services, client-side stacks, and regulation. Looking ahead to 2025, the central technical trends will continue to revolve around hardware acceleration, especially FPGA, network optimization through both physical proximity and lean protocols, and continued improvement in software and algorithmic efficiency.

Core technical path and considerations:

  1. Exchange layer:
    • Trading units and flow rate: Sufficient trading-unit flow rate is a prerequisite for smooth entry into the exchange. For HFT and low-latency traders, purchasing additional flow rate is almost unavoidable. However, exchange controls on submission speed, especially for HFT, mean that chasing unlimited throughput is not feasible. Optimization must happen within the compliance framework.
    • Co-location: Deploying trading servers inside exchange data centers is the foundation of minimum network latency. Rack position, cabinet specification, and internal cross-connect design all matter.
  2. Broker layer:
    • High-speed counters: Using technologically advanced counters, based on FPGA, LDP, UFT, and similar architectures, can materially reduce broker-internal latency. It is important to focus on real end-to-end performance, not only internal benchmarks, and also to consider feature completeness.
    • DMA: Where the broker offers high-quality DMA, it can effectively bypass parts of the broker’s internal stack and is an effective path to lower latency. The cleanliness and stability of the DMA path should be evaluated carefully.
    • Integrated broker low-latency solutions: Leading brokers are investing heavily in in-house or heavily customized low-latency platforms such as CITIC’s CATS, often combining market data, trading, and risk control in an optimized architecture.
  3. Client and supporting technology layer:
    • Market data: Access to exchange Level-2 depth data, together with FPGA decoding and binary protocols, is a prerequisite for fast decision-making.
    • Network optimization: Use leased lines, optimize internal topology, and deploy low-latency network hardware and NICs with SR-IOV or kernel bypass.
    • System and hardware tuning: Tune server CPUs, memory, and the operating system carefully, including core pinning, disabling power saving, using fast memory, and applying real-time patches.
    • Algorithms and software: The trading algorithm itself should be efficient and avoid unnecessary computation. Compiled languages such as C++ are preferable for execution-critical paths.

Key influences and trends:

  • Latency is additive: End-to-end latency is the sum of delays across all links. Extreme optimization in one stage can easily be canceled out by bottlenecks elsewhere, so a full-path coordinated approach is mandatory.
  • Cost-benefit trade-off: Lower latency almost always requires more capital and engineering investment. Institutions should assess carefully how latency-sensitive their own strategies really are.
  • Continued regulatory evolution: The program-trading rules effective from 2024 create new behavioral boundaries for the market. Regulators may further refine controls on HFT and abnormal trading and may adjust fee schedules. Compliance will remain an inseparable part of low-latency trading.

Strategic recommendations for 2025:

  1. Form an integrated low-latency strategy: Treat exchange access, broker selection, technical architecture, algorithmic optimization, and compliance management as one unified plan.
  2. Prioritize physical proximity: Use exchange co-location wherever possible. It remains the most effective way to reduce network latency.
  3. Choose brokers and technical partners carefully: Evaluate broker investment and capability in high-speed counters, DMA, and hosting resources, as well as the real performance and support quality of vendor products.
  4. Invest in core technical capability: Institutions with sufficient resources should consider FPGA and similar hardware acceleration in critical links such as market-data handling, core trading logic, and risk control, and may choose to build some core components in-house.
  5. Optimize continuously: Low-latency technology evolves quickly, so firms should establish ongoing monitoring, evaluation, and optimization of trading-system performance.
  6. Stay highly attentive to regulation and adapt proactively: Build the latest reporting, HFT classification, and abnormal-trading-monitoring rules into both system design and daily operations so that speed does not come at the expense of compliance or resilience.

In short, the path to minimum-latency trading in China’s A-share market in 2025 will be defined by a dual emphasis on technical innovation and regulatory discipline. Only those participants who understand the technological frontier deeply, read market structure accurately, and comply rigorously with the regulatory framework will be able to hold an advantage in this contest of speed and intelligence.

Works cited

  1. A Review of the Impact of High-Frequency Trading on the Securities Market, accessed May 31, 2025, http://docs.static.szse.cn/www/aboutus/research/secuities/daily/W020190725498844424029.pdf
  2. The Past and Present of High-Frequency Trading, Finally Explained Clearly, accessed May 31, 2025, https://wallstreetcn.com/articles/3642103
  3. Regulation of Unfair High-Frequency Trading in the United States and Its Implications for Securities-Law Reform, accessed May 31, 2025, https://docs.static.szse.cn/www/aboutus/research/secuities/daily/W020180829345937001517.pdf
  4. Executing Trades Online | J.P. Morgan Private Bank Asia, accessed May 31, 2025, https://privatebank.jpmorgan.com/apac/cn/about-us/digital-capabilities/how-to-pages/execute-trades-online
  5. Full-Process Management Solution for Sales Order Execution, accessed May 31, 2025, https://hc.jiandaoyun.com/solution/17341
  6. Shanghai Stock Exchange Trading Rules - Securities - ICBC China, accessed May 31, 2025, https://www.icbc.com.cn/icbc/%E7%BD%91%E4%B8%8A%E8%82%A1%E7%A5%A8/%E6%8A%95%E8%B5%84%E8%80%85%E6%95%99%E8%82%B2/%E6%B3%95%E8%A7%84%E6%9D%A1%E4%BE%8B/%E4%B8%8A%E6%B5%B7%E8%AF%81%E5%88%B8%E4%BA%A4%E6%98%93%E6%89%80%E4%BA%A4%E6%98%93%E8%A7%84%E5%88%99123.htm
  7. Secondary-Market Trading Mechanism: Entrusted Trading, accessed May 31, 2025, http://www.csrc.gov.cn/csrc/c100211/c6288926/6288926/files/f2a0e5fdf2db4e00afb4c6ee60a245d8.pdf
  8. Shanghai Stock Exchange: Home Page, accessed May 31, 2025, http://www.sse.com.cn
  9. Shanghai Stock Exchange Trading Rules, accessed May 31, 2025, https://www.htsec.com/jfimg/colimg/upload/20151026/65891445823853588.pdf
  10. docs.static.szse.cn, accessed May 31, 2025, https://docs.static.szse.cn/www/aboutus/trends/news/W020230201595730281095.pdf
  11. Understanding the BSE Stock Trading Rules in One Article, accessed May 31, 2025, http://edu.eastmoneysec.com/a/202110242152225529.html
  12. Interpretation of BSE Stock Trading Rules, accessed May 31, 2025, https://www.investor.org.cn/learning_center/gmjytx/yjs/kj/jyxl_3472/202303/P020230307532838963458.pdf
  13. A New Benchmark in Ultra-Fast Trading: Huatai Securities Launches China’s First FPGA Securities Trading Counter, accessed May 31, 2025, https://www.yicai.com/news/100827548.html
  14. White Paper on Securities-Market Acceleration Card Technology, accessed May 31, 2025, https://ehiway.cn/uploads_file/20220815/20220815163104_303.pdf
  15. 2023 Low-Latency Laboratory Research Requirement List, accessed May 31, 2025, https://www.sse.com.cn/aboutus/purchase/information/c/10749780/files/83541bddd25d4952ba9e6d2e2bcdfe84.pdf
  16. Low-Latency Solution for High-Frequency Trading in the Securities Industry, accessed May 31, 2025, https://www.sangfor.com.cn/industry/financial/securities/low-latency-solution
  17. Infrastructure Optimization for High-Frequency Trading | Official AWS Blog, accessed May 31, 2025, https://aws.amazon.com/cn/blogs/china/building-cefi-high-frequency-trading-system-on-aws/
  18. AMD Launches an FPGA-Based Accelerator Purpose-Built for Ultra-Low-Latency Electronic Trading, accessed May 31, 2025, https://www.amd.com/zh-cn/newsroom/press-releases/2023-9-27-amd-unveils-purpose-built-fpga-based-accelerator-.html
  19. CITIC Securities Automated Trading Platform, accessed May 31, 2025, http://cats.citicsinfo.com:22000/cats
  20. Notice of the General Office of the State Council Forwarding the Opinions on Strengthening Regulation, Preventing Risk, and Promoting High-Quality Development of the Futures Market, accessed May 31, 2025, http://www.csrc.gov.cn/csrc/c100028/c7511274/content.shtml
  21. Measures for the Administration of Securities Exchanges - China Government Website, accessed May 31, 2025, https://www.gov.cn/zhengce/2021-12/16/content_5724562.htm
  22. The CSRC’s Most Candid Statement on Quant Trading Yet: How Should It Be Viewed and Regulated?, accessed May 31, 2025, https://www.mszq.com/cms/article/9503
  23. SSE Q&A on the Formal Release of the Program-Trading Implementation Rules and the Solicitation of Opinions on Supporting Business Rules, accessed May 31, 2025, https://www.sse.com.cn/aboutus/mediacenter/hotandd/c/c_20250403_10776805.shtml
  24. Shanghai, Shenzhen, and Beijing Exchanges Release Program-Trading Implementation Rules, accessed May 31, 2025, https://jrj.sh.gov.cn/SC212/20250407/c291f4a2534947d7ae48769c3e223d1b.html
  25. CSRC Officially Releases the Administrative Provisions for Program Trading in the Securities Market (Trial), accessed May 31, 2025, https://www.junhe.com/legal-updates/2429
  26. np-newsimg.dfcfw.com, accessed May 31, 2025, https://np-newsimg.dfcfw.com/download/A2_cms_f_20250403183112931337?direct=1&abc8962.pdf
  27. SZSE ChiNext Listed Company Self-Regulatory Guide No. 1: Business Handling (Revised December 2023), accessed May 31, 2025, http://www.csf.com.cn/publish/main/1040/1043/20240524151520092597256/1716534934251.pdf
  28. SZSE Implementation Rules for Program Trading Management, accessed May 31, 2025, https://np-newsimg.dfcfw.com/download/A2_cms_f_20250403183112931337?direct=1&abc8962.pdf?
  29. BSE Q&A on the Formal Release of the Program-Trading Implementation Rules, accessed May 31, 2025, https://www.bseinfo.net/important_news/200025377.html
  30. Drafting Notes for the SZSE Implementation Rules for Program Trading Management, accessed May 31, 2025, https://docs.static.szse.cn/www/aboutus/trends/news/W020250403603802336330.pdf
  31. Shanghai, Shenzhen, and Beijing Exchanges Introduce Self-Regulatory Rules to Advance Program-Trading Oversight, accessed May 31, 2025, https://wap.eastmoney.com/a/202504033364930057.html
  32. Supplemental Notice on Adjusting Participant Trading-Business-Unit Flow-Rate Configuration Functions - SSE, accessed May 31, 2025, https://www.sse.com.cn/lawandrules/sselawsrules/trade/universal/c/c_20210128_5312096.shtml
  33. baike.eastmoney.com, accessed May 31, 2025, https://baike.eastmoney.com/item/%E4%BA%A4%E6%98%93%E5%8D%95%E5%85%83#:~:text=%E4%BA%A4%E6%98%93%E5%8D%95%E5%85%83%E6%98%AF%E6%8C%87%E4%BC%9A%E5%91%98,%E4%BA%A4%E6%98%93%E5%8D%95%E5%85%83%E7%9A%84%E4%BD%BF%E7%94%A8%E6%9D%83%E3%80%82
  34. Rules Governing Trading Units on Chinese Stock Exchanges - Eastmoney, accessed May 31, 2025, https://baike.eastmoney.com/item/%E4%BA%A4%E6%98%93%E5%8D%95%E5%85%83
  35. docs.static.szse.cn, accessed May 31, 2025, http://docs.static.szse.cn/www/disclosure/notice/W020180328432928783546.pdf
  36. Frequently Asked Questions | Shanghai Stock Exchange One-Stop Service, accessed May 31, 2025, https://one.sse.com.cn/member/hyjgzj/hy_hyzc_cjwt/
  37. www.psbc.com, accessed May 31, 2025, https://www.psbc.com/cn/grfw/tzlc/zjcg/flfga/202010/P020210104677709474406.pdf
  38. SZSE Member Management Rules - ICBC, accessed May 31, 2025, https://www.icbc.com.cn/icbc/%E7%BD%91%E4%B8%8A%E8%82%A1%E7%A5%A8/%E6%8A%95%E8%B5%84%E8%80%85%E6%95%99%E8%82%B2/%E6%B3%95%E8%A7%84%E6%9D%A1%E4%BE%8B/%E6%B7%B1%E5%9C%B3%E8%AF%81%E5%88%B8%E4%BA%A4%E6%98%93%E6%89%80%E4%BC%9A%E5%91%98%E7%AE%A1%E7%90%86%E8%A7%84%E5%88%99123.htm
  39. Measures for the Administration of Securities Exchanges (2021 Revision) - Peking University Law, accessed May 31, 2025, https://law.pkulaw.com/chinalaw/76f013fed31e1c84bdfb.html
  40. National Equities Exchange and Quotations Mobile Site - Trading Supervision, accessed May 31, 2025, https://www.neeq.cc/m/service/jyjg.html
  41. CSRC Order No. 204: Measures for the Administration of Securities Brokerage Business - China Government Website, accessed May 31, 2025, https://www.gov.cn/gongbao/content/2023/content_5750630.htm
  42. Measures for the Administration of Securities Exchanges - China Securities Regulatory Commission, accessed May 31, 2025, http://www.csrc.gov.cn/csrc/c106256/c1653990/content.shtml
  43. Trading Units | SSETech, accessed May 31, 2025, https://www.ssetech.com.cn/product/exchange/tradingUnit/
  44. Open-Service Platform - Shanghai Stock Exchange, accessed May 31, 2025, https://www.sse.com.cn/transparency/services/
  45. Trusted Data Services - Shenzhen Securities Information Co., accessed May 31, 2025, http://www.szsi.cn/cpfw/sjfw/
  46. SZSE Consolidated Transaction Service Access Guide, accessed May 31, 2025, https://docs.static.szse.cn/www/marketServices/technicalservice/notice/W020190822521412095060.pdf
  47. “Shenwan Hongyuan Securities” on the App Store, accessed May 31, 2025, https://apps.apple.com/cn/app/%E7%94%B3%E4%B8%87%E5%AE%8F%E6%BA%90%E8%AF%81%E5%88%B8/id914460168
  48. What Is a Trading Seat? - Eastmoney Encyclopedia Entry, accessed May 31, 2025, https://baike.eastmoney.com/item/%E4%BA%A4%E6%98%93%E5%B8%AD%E4%BD%8D
  49. baike.eastmoney.com, accessed May 31, 2025, https://baike.eastmoney.com/item/%E4%BA%A4%E6%98%93%E5%B8%AD%E4%BD%8D#:~:text=%EF%BC%A1%E8%82%A1%E5%B8%AD%E4%BD%8D%E6%98%AF%E7%BB%8F%E8%90%A5,%E5%B8%AD%E4%BD%8D%E5%92%8C%E8%87%AA%E8%90%A5%E5%B8%AD%E4%BD%8D%E3%80%82
  50. Program Trading and High-Frequency Trading: History, Present, and Future, accessed May 31, 2025, http://sf.ruc.edu.cn/publish/sf/alk/alfl/jr/359ddb69799d45d6964814cf521fcee8.htm
  51. Products and Services | SSETech, accessed May 31, 2025, https://www.ssetech.com.cn/product/
  52. SSETech Zhengtong Cloud Product and Service White Paper, accessed May 31, 2025, https://www.sseinfo.com/home/news/c/10010563/files/6f4c0e438925418f815155b7a22b7879.pdf
  53. SSETech Zhengtong Cloud Product and Service White Paper, accessed May 31, 2025, https://www.ssetech.com.cn/news/notice/c/10766138/files/598248ec2c8046d4a65e98f77d46c9c2.pdf
  54. Data-Center Business - Shenzhen Securities Communication Co., accessed May 31, 2025, https://www.sscc.com/business_system/data_center/
  55. Shenzhen Securities Communication Co. - Home, accessed May 31, 2025, https://www.sscc.com/
  56. Administrative Provisions on Co-Location in the Securities and Futures Market (Exposure Draft) - CSRC, accessed May 31, 2025, http://www.csrc.gov.cn/csrc/c101981/c7403869/7403869/files/%E9%99%84%E4%BB%B61%EF%BC%9A%E3%80%8A%E8%AF%81%E5%88%B8%E6%9C%9F%E8%B4%A7%E5%B8%82%E5%9C%BA%E4%B8%BB%E6%9C%BA%E4%BA%A4%E6%98%93%E6%89%98%E7%AE%A1%E7%AE%A1%E7%90%86%E8%A7%84%E5%AE%9A%EF%BC%88%E5%BE%81%E6%B1%82%E6%84%8F%E8%A7%81%E7%A8%BF%EF%BC%89%E3%80%8B.pdf
  57. Wangsu: Global Service Provider for CDN, Edge Computing, Cloud Security, Zero Trust, SASE, WAAP, and Liquid-Cooled Information Infrastructure, accessed May 31, 2025, https://www.wangsu.com/
  58. Bank of China Global Portal Home Page, accessed May 31, 2025, https://www.boc.cn/
  59. Connecting a Local IDC to a Cloud VPC via a Dedicated Physical Line - Alibaba Cloud Documentation, accessed May 31, 2025, https://help.aliyun.com/zh/express-connect/getting-started/connect-a-data-center-to-a-vpc-by-using-an-express-connect-circuit
  60. Access Modes and Solution Comparison for Express Connect - Alibaba Cloud Documentation, accessed May 31, 2025, https://help.aliyun.com/zh/express-connect/product-overview/functions-and-features
  61. Announcement on the Reply to the Inquiry Letter from the Shenzhen Stock Exchange by Net263 Ltd., accessed May 31, 2025, http://static.cninfo.com.cn/finalpage/2018-06-26/1205090160.PDF
  62. Financial User Case Study: Building the Cloud Foundation of a CTP Futures Trading System with Hyper-Convergence - SmartX, accessed May 31, 2025, https://www.smartx.com/blog/2023/11/finance-futures-ctp/
  63. Hundsun PBOX High-Speed Trading System - Xiangcai Securities, accessed May 31, 2025, https://www.xcsc.com/main/privateplacementservice/gsssystem/quicktradingrunway/index.shtml
  64. Industrial Securities SMT Portal, accessed May 31, 2025, https://smt.xyzq.com.cn/
  65. In-Depth Analysis of Hundsun Electronics (600570.SH), accessed May 31, 2025, https://pdf.dfcfw.com/pdf/H3_AP202105271494331362_1.pdf
  66. Software Downloads - Changjiang Securities, accessed May 31, 2025, https://www.95579.com/main/software/index.shtml
  67. New Core IT Direction in the Securities Industry: Distributed Architecture plus Domestic Innovation, accessed May 31, 2025, https://pdf.dfcfw.com/pdf/H3_AP202308071593395381_1.pdf
  68. Kingdom’s KFOS 3.5 High-Speed Trading System Officially Launches, accessed May 31, 2025, https://www.szkingdom.com/news/606
  69. The New Hundsun High-Speed PB Trading System Has Arrived!, accessed May 31, 2025, https://www.hs.net/openplat-doc/informationDetail/detail/788
  70. XELE-TRADE® - AcceleCom Information Technology Co., accessed May 31, 2025, https://www.accelecom.com/products/xele-trade/
  71. Huasheng Ultra-Low-Latency Trading System with Microsecond Delay, accessed May 31, 2025, https://www.vbkr.com/cn/solve/fast-trade
  72. CITIC Securities Institutional Services, accessed May 31, 2025, https://i.citics.com/
  73. CITIC Securities Quantitative Trading Platform (CATS), accessed May 31, 2025, http://cats.citicsinfo.com:22000/cats/
  74. Quantitative Trading - Guotai Junan Futures, accessed May 31, 2025, https://www.gtjaqh.com/mb/c/quantitative
  75. Business Introduction to Quantitative Trading - Guotai Junan Futures, accessed May 31, 2025, https://www.gtjaqh.com/mb/c/S0uGzfU
  76. Asset Management, Market Research, Fund Administration, and Related Services - China Merchants Securities, accessed May 31, 2025, https://www.newone.com.cn/newonefront/gjyw.html
  77. China Merchants Securities Clients Try the Industry’s First Quant Trading Platform Service - PR Newswire, accessed May 31, 2025, https://cnmobile.prnasia.com/story/64054-1.shtml
  78. Huaxin Securities Welcomes You, accessed May 31, 2025, https://www.cfsc.com.cn/www/html/154/2020-06-22/280.html?v=1629968964105
  79. Custody and Outsourcing Services - China Merchants Securities Official Website, accessed May 31, 2025, https://www.newone.com.cn/newonefront/zctg.html
  80. China Merchants Securities Official Website, accessed May 31, 2025, https://www.cmschina.com/
  81. “China Galaxy Securities - Fast Account Opening” on the App Store, accessed May 31, 2025, https://apps.apple.com/cn/app/%E4%B8%AD%E5%9B%BD%E9%93%B6%E6%B2%B3%E8%AF%81%E5%88%B8-%E6%9E%81%E9%80%9F%E5%BC%80%E6%88%B7/id441890370
  82. Galaxy Securities Dazhihui 4+ - App Store, accessed May 31, 2025, https://apps.apple.com/cn/app/%E9%93%B6%E6%B2%B3%E8%AF%81%E5%88%B8%E5%A4%A7%E6%99%BA%E6%85%A7/id1598550039
  83. “Shenwan Futures - Fast, Professional, and Compliant Account Opening 17+” on the App Store, accessed May 31, 2025, https://apps.apple.com/cn/app/%E7%94%B3%E4%B8%87%E6%9C%9F%E8%B4%A7-%E6%9E%81%E9%80%9F%E5%BC%80%E6%88%B7%E4%B8%93%E4%B8%9A%E6%AD%A3%E8%A7%84/id1207999864
  84. About Us - CITIC Futures International, accessed May 31, 2025, https://www.citicsf.com.hk/about/
  85. Scenario Solutions - Kingdom, accessed May 31, 2025, https://www.szkingdom.com/fangan/57
  86. Comparative Table for the Revised Model Risk Disclosure Statement for DMA Trading over Dedicated Lines, accessed May 31, 2025, https://www.twsa.org.tw/D01/doc/%E5%B0%88%E5%B1%AC%E7%B7%9A%E8%B7%AFDMA%E4%B8%8B%E5%96%AE%E4%BA%A4%E6%98%93%E5%A7%94%E8%A8%97%E9%A2%A8%E9%9A%AA%E9%A0%90%E5%91%8A%E6%9B%B8%E7%AF%84%E6%9C%AC.pdf
  87. Shanghai Futures Exchange Second-Generation Market Data Release Platform Interface Specification, accessed May 31, 2025, https://www.shfe.com.cn/services/technology/technical_download/201906/P020240320683540012629.pdf
  88. Technical White Paper for Trading and Related Systems of the Shanghai Stock Exchange (Revised August 2023), accessed May 31, 2025, https://www.sse.com.cn/services/tradingtech/services/c/10649698/files/b66e6432ef5a42e1b1124ff626694258.docx
  89. Description of SZSE Internet Market-Data Access Service, accessed May 31, 2025, http://www.szsi.cn/cpfw/fwsq/hq/%E6%B7%B1%E4%BA%A4%E6%89%80%E8%A1%8C%E6%83%85%E4%BA%92%E8%81%94%E7%BD%91%E6%8E%A5%E5%85%A5%E6%9C%8D%E5%8A%A1%E8%AF%B4%E6%98%8E.pdf
  90. Shenzhen Market Data Authorization - Shenzhen Securities Information Co., accessed May 31, 2025, http://www.szsi.cn/cpfw/fwsq/hq/yw-2.htm
  91. Directory of Open-Source Vendors - China Academy of Information and Communications Technology, accessed May 31, 2025, http://www.caict.ac.cn/xwdt/ynxw/202105/P020210519482142667760.pdf
  92. Cross-Border STEP Market Data Interface Specification - Shenzhen Stock Exchange, accessed May 31, 2025, https://biz.sscc.com/fileDownload/downloadOutFile?id=1187
  93. SZSE Internet Market-Data Access Service - SSE LDDS Market-Data Interface Specification, accessed May 31, 2025, http://www.szsi.cn/cpfw/fwsq/hq/%E6%B7%B1%E4%BA%A4%E6%89%80%E8%A1%8C%E6%83%85%E4%BA%92%E8%81%94%E7%BD%91%E6%8E%A5%E5%85%A5%E6%9C%8D%E5%8A%A1-%E4%B8%8A%E4%BA%A4%E6%89%80LDDS%E8%A1%8C%E6%83%85%E6%8E%A5%E5%8F%A3%E8%AF%B4%E6%98%8E%E4%B9%A6.pdf
  94. SZSE Internet Market-Data Access Service Interface Specification, accessed May 31, 2025, http://www.szsi.cn/cpfw/fwsq/hq/%E6%B7%B1%E4%BA%A4%E6%89%80%E8%A1%8C%E6%83%85%E4%BA%92%E8%81%94%E7%BD%91%E6%8E%A5%E5%85%A5%E6%9C%8D%E5%8A%A1%E6%8E%A5%E5%8F%A3%E8%AF%B4%E6%98%8E%E4%B9%A6.pdf
  95. SSE Low-Latency Market Data Distribution System (LDDS) Interface Specification, accessed May 31, 2025, https://www.sseinfo.com/services/assortment/market/hqywwd/wdkfcsjk/c/10764835/files/a56ccfba62ee4507babe14a0a4646d5e.pdf
  96. Domestic Information Vendors - Shanghai Stock Exchange, accessed May 31, 2025, https://www.sse.com.cn/services/information/vendors/
  97. SSE LDDS Auction Level-2 Market-Data Interface Specification - SSE Infonet Co., accessed May 31, 2025, https://www.sseinfo.com/services/assortment/market/hqywwd/wdkfcsjk/c/10764834/files/8f5432f88c904455828b4804b9ae03d6.pdf
  98. Non-Display Market Data - FitC, accessed May 31, 2025, https://www.dfitc.com.cn/dfitc/scfw/xqsjfw/fzsxq/index.html
  99. China Investment Information Services Ltd. – Document Download Area, accessed May 31, 2025, https://www.ciis.com.hk/hongkong/sc/productsandservices/documentdownloads/index.shtml
  100. biz.sscc.com, accessed May 31, 2025, https://biz.sscc.com/fileDownload/downloadOutFile?id=1057
  101. IS120 SSE Market Data Gateway BINARY Interface Specification Version 0.588, accessed May 31, 2025, https://www.sse.com.cn/services/tradingtech/development/c/10753973/files/148f0c9b92204c9190fce0b17f53fcc5.pdf
  102. Derivatives Platform - Singapore Exchange, accessed May 31, 2025, https://www.sgx.com/zh-hans/derivatives/platforms
  103. FIX API: Financial Information eXchange Protocol - Swissquote, accessed May 31, 2025, https://www.swissquote.com/cn-row/private/trade/platforms/forex-cfds/fix-api
  104. Frontiers of Trading Technology - Shanghai Stock Exchange, accessed May 31, 2025, http://www.sse.com.cn/services/tradingtech/transaction/ITRDC_TechMag_040_202011.pdf
  105. Initial Public Offering Prospectus of Net263 Ltd. - Sina Finance, accessed May 31, 2025, https://vip.stock.finance.sina.com.cn/corp/view/vISSUE_RaiseExplanationDetail.php?stockid=002467&id=598830
  106. Shenzhen Securities Communication Network Access Specification - Shenzhen Stock Exchange, accessed May 31, 2025, https://biz.sscc.com/fileDownload/downloadOutFile?id=1231
  107. CN110460540A - A Low-Latency High-Frequency Trading System and Method - Google Patents, accessed May 31, 2025, https://patents.google.com/patent/CN110460540A/zh
  108. Citrix SD-WAN WANOP 11.0 - Product Documentation, accessed May 31, 2025, https://docs.netscaler.com/zh-cn/citrix-sd-wan-wanop/11/citrix-sd-wan-wanop-11.0.pdf
  109. AlphaDesk - Cloud-Based Order and Portfolio Management - London Stock Exchange Group (LSEG), accessed May 31, 2025, https://www.lseg.com.cn/zh/data-analytics/trading-solutions/alphadesk-order-management
  110. Shenwan Futures High-Frequency and Program Trading System - LeFile, accessed May 31, 2025, http://dcg.app.lefile.cn/Public/pdf/94.pdf
  111. Investment Research System - Kingdom, accessed May 31, 2025, https://www.szkingdom.com/fangan/54
  112. Administrative Provisions for Program Trading in the Securities Market (Trial) (CSRC Announcement [2024] No. 8) - Zhengzhou Foreign Enterprise Service Center, accessed May 31, 2025, https://www.waizi.org.cn/doc/253221.html
  113. CSRC Releases the Administrative Provisions for Program Trading in the Securities Market (Trial) - Beijing Municipal Financial Office, accessed May 31, 2025, https://jrj.beijing.gov.cn/jrgzdt/202405/t20240521_3690224.html
  114. accessed January 1, 1970, https://www.csrc.gov.cn/csrc/c100028/c7438108/content.shtml
  115. Key Regulatory Requirements for Algorithmic Trading in Major Overseas Financial Markets - Asset Management Association of China, accessed May 31, 2025, https://www.amac.org.cn/hyyj/sy/202312/P020231228607112393726.pdf

Research Report on Technical Implementation Paths for Low-Latency Trading in China's A-Share Market in 2025

https://en.heth.ink/AShareLowLatency/

Author

YK

Posted on

2025-06-02

Updated on

2025-06-02

Licensed under

Categories

Recents

×