Our System Architecture

Architectural Overview

Core Architectural Principles

• Performance-First Design: Optimized for ultra-low latency in critical trading paths
• Distributed Processing: Microservices with event-driven communication
• Scalable Components: Independently scalable services to handle varying loads
• Resilient Operations: No single points of failure across the system
• Comprehensive Observability: Real-time monitoring and diagnostics

Touch-Fire's system architecture represents a careful balance between modern cloud-native principles and the stringent performance requirements of financial trading systems. We've created a hybrid approach that delivers the scalability and resilience of cloud technologies while maintaining the ultra-low latency required for high-frequency trading.

Architectural Transformation

From 2019 to 2022, Touch-Fire underwent a major architectural transformation to embrace cloud-native principles while preserving our market-leading latency.

This journey was triggered by the need to support both high-frequency trading clients and retail-facing crypto exchange platforms with thousands of account opening requests per hour and tens of thousands of users requiring price information.

The result is a best-of-both-worlds solution that combines the reliability and scalability of modern cloud architectures with the performance characteristics required for financial markets.

Database Architecture

Breaking Down the Monolith

A critical part of our architectural transformation was breaking down our monolithic database into purpose-specific data stores:

• In-Memory Distributed Database: For real-time metadata in latency-sensitive services
• PostgreSQL Database: For configuration, transaction management, and reporting
• InfluxDB Time Series Database: For our real-time OHLC bar factory and market data

This approach has dramatically improved developer productivity, system scalability, and operational flexibility while reducing dependencies on expensive proprietary database systems.

Digital Twin Observability

Our system monitoring is based on a digital twin approach to observability, using OpenTelemetry standards, service mesh integration, and self-registration of components:

• Comprehensive Monitoring: Real-time visibility into the entire system
• Unified Management Console: Integrated application and infrastructure controls
• Centralized Logging: Aggregated logs with advanced search and correlation
• Distributed Tracing: End-to-end transaction visibility across services
• Automated Failover: Instant failover of stateful services based on digital twin state

This approach provides unprecedented visibility into system behavior and enables rapid problem diagnosis, performance optimization, and automated remediation.

Time Management and Scheduling

Financial markets operate across multiple time zones, calendars, and session types. Our architecture includes sophisticated time management capabilities:

From Batch to Stream Processing

Continuous Event Processing

Traditional financial systems rely heavily on end-of-day batch processing. Our architecture is transitioning to continuous stream processing for several key advantages:

• Real-Time Updates: Immediate position and risk calculations
• Higher Volume Capacity: Scale to handle millions of events per second
• System-Wide Events: Instant propagation of status changes like margin warnings
• Reduced Processing Windows: Elimination of nighttime batch processing constraints
• Immediate Alerting: Real-time notification of critical conditions

This shift to stream processing is particularly important for high-frequency trading clients who require immediate feedback on positions, risk, and market conditions.

DevOps and Continuous Delivery

Our architectural transformation included a major evolution in our development and deployment practices: