Guotai Haitong Securities is a long-standing leader in China's capital market. The company serves over 40 million retail customers and maintains a top-tier position in the industry, underscoring its pivotal role in the financial ecosystem.
As one of China's largest and most influential financial service providers, Guotai Haitong must provide its vast retail customer base with comprehensive financial data services, including real-time market data for A-share and Hong Kong stock markets.
Challenge
The challenge arose from the extreme scale and strict performance requirements of pushing real-time market data to millions of clients. Specifically:
- Massive Concurrency: During peak trading hours, a single popular stock could have hundreds of thousands of concurrent subscriptions. The system needed to reliably handle millions of concurrent connections.
- Ultra-Low Latency: Market data latency is critical for user decision-making. The end-to-end latency for pushing stock quotes had to be less than tens of milliseconds.
- Extreme One-to-Many Distribution: The system had to efficiently distribute a single market data update to hundreds of thousands of subscribers simultaneously, pushing traditional messaging systems beyond their limits.
- High Availability: As a core financial infrastructure, the system demanded continuous operation and high reliability without any single point of failure, even under massive loads.
Traditional systems often faced issues like slow responses, message backlog, and instability when confronting this scale of concurrent subscriptions and low-latency demands. The company needed to upgrade to a robust, high-performance infrastructure for real-time message distribution.
Solution
EMQX provided a high-performance, real-time market data push system leveraging a cloud-native, distributed cluster. This solution was designed to handle the extreme concurrency and ultra-low latency required by the securities trading environment.
Key Features of the Solution:
- High-Availability Cluster: A multi-node, high-availability EMQX cluster was deployed, with each node configured for optimal performance (e.g., 64-core CPU, 128GB memory, 32Gbps network bandwidth).
- Extreme Distribution Optimization: Through fine-tuned configurations, the cluster achieved a one-to-many message push throughput of 5,000 messages/second per core. This optimization enabled the system to efficiently distribute a single market data update to 500,000 subscribing clients.
- Elastic Scalability: The cluster utilizes a Core + Replicant deployment mode, supporting elastic scaling to easily handle sudden traffic surges.
- Asynchronous Shard Scheduling: EMQX implemented a custom-optimized version with asynchronous shard scheduling to significantly boost message push efficiency.
Project Outcomes
The EMQX-based solution successfully achieved unprecedented performance and enhanced business value:
- Performance Breakthroughs: The system supports hundreds of thousands of concurrent subscriptions to a single stock, achieving an average end-to-end latency in tens of milliseconds, significantly surpassing business requirements.
- Resource Efficiency: The 4-node cluster demonstrated exceptional efficiency; at high throughput (hundreds of thousands of messages/second), CPU utilization remains below 10% and memory usage is about 12%.
- Enhanced Competitiveness: By delivering stable and highly efficient real-time market data, the system significantly improved user experience for 40 million retail customers. The scalable and reliable architecture ensures the stability of the trading system and supports future business growth, reinforcing Guotai Haitong's market leadership.