High-frequency trading systems process 50,000+ messages per second with latency budgets measured in microseconds. In this environment, a single 50ms garbage collection pause means missed trades and violated SLAs.


This webinar walks through TradeStreamEE, a reference architecture demonstrating how Jakarta EE on Payara Micro achieves HFT-grade performance when combined with Azul Platform Prime's C4 garbage collector.

The Problem: Why G1GC's stop-the-world pauses create unacceptable latency in financial systems, and how pause times scale with heap size

Architecture: Zero-copy message pipeline using Aeron IPC for shared-memory transport and SBE for binary encoding—eliminating allocation in the hot path

Stress Testing: Six targeted GC stress scenarios that expose specific collector weaknesses: steady load, growing heap, promotion storms, fragmentation, and cross-generational references

Results: Direct comparison of G1GC vs C4 under identical workloads. In PROMOTION_STORM testing (300 MB/s, 50% survival), G1 showed 20 SLA violations with 25ms maximum pause, while C4 delivered zero violations with 2ms maximum pause

Deployment: Running the application, monitoring GC behavior via JMX, and applying stress scenarios via REST API

Attendees will see the live dashboard visualizing real-time GC pause percentiles and learn why enterprise Java can compete with C++ in latency-sensitive workloads- when paired with the right runtime.

High-Frequency Trading on Jakarta EE: GC Stress Testing with Azul C4 & Payara Micro