FakeX

FakeX is an experimental trading engine built to explore order matching, risk checks, settlement, write-ahead logging, and low-latency event processing in Java.

The project is intentionally small and is not intended to be a production-ready exchange. Its purpose is to demonstrate the core ideas behind a trading system and provide a foundation for further experiments and performance testing.

Current Features

• Buy and sell limit orders
• Price-time priority matching
• Partial fills and matching against multiple orders
• Order cancellation
• Basic account and balance management
• Basic buy-side funds reservation
• Separate risk, matching, and settlement processing stages
• Command processing through LMAX Disruptor
• Command journaling with Chronicle Queue
• JMH benchmarks

Architecture

Commands are published to a single-producer Disruptor ring buffer and processed through several stages:

Command ├── CommandLogger └── RiskEngine ↓ MatchingEngine ↓ SettlementEngine

The order book currently uses:

• PriorityQueue for buy and sell orders
• Long2ObjectHashMap for lookup by order ID

This implementation favors simplicity and readability over a fully optimized order-book data structure.

Limitations

FakeX is a prototype and currently has several intentional limitations:

• A single order book
• Only limit orders
• Simplified asset and balance model
• Incomplete reservation handling
• No persisted state recovery or journal replay
• No snapshots
• No networking or exchange protocol
• No market-data feed
• No high availability or replication
• Cancellation from a priority queue is linear-time

Requirements

• JDK 23 or newer
• Gradle, or the included Gradle wrapper

Build

./gradlew build

Run

./gradlew run

Benchmarks

The project includes JMH benchmarks for selected engine operations.

Run all benchmarks:

./gradlew jmh

Run a specific benchmark:

./gradlew jmh --includes=BasicOrderMatchingBenchmark

Benchmark results should be treated as local experimental measurements rather than production performance claims.

Main Libraries

• LMAX Disruptor
• Agrona
• Chronicle Queue
• Chronicle Wire
• JMH

Possible Next Steps

• Deterministic command replay
• Correct reservation release on cancellation and partial fills
• Sell-side asset reservation
• Atomic risk, matching, and settlement transitions
• Multiple instruments
• A more efficient order-book representation
• Snapshots and recovery
• Property-based and model-based tests
• More realistic latency and throughput benchmarks