1. What Is High-Frequency Trading?
High-Frequency Trading is the use of automated systems to trade financial instruments at microsecond or nanosecond latencies.
Key characteristics:
- Extremely low latency
- High message throughput
- Short holding periods
- Strong focus on infrastructure
HFT is a systems engineering problem, not a pure finance problem.
2. The Trading Lifecycle
Every trade follows a deterministic pipeline:
Each stage has a latency budget.
Missing the budget = losing the trade.
3. Market Data Feeds
Exchanges publish:
- Order book updates
- Trades
- State changes
Feeds are:
- High volume
- Burst-heavy
- Often UDP multicast
Processing market data fast is more important than strategy sophistication.
4. Strategy Layer
Strategies operate on:
- Order book state
- Derived signals
- Statistical models
Constraints:
- Must be deterministic
- Must avoid allocations
- Must be predictable
In HFT, simplicity beats complexity.
5. Risk Management (Inline)
Risk checks happen:
- Before every order
- In microseconds
Examples:
- Position limits
- Price bands
- Message rate limits
Risk is enforced in the hot path.
6. Order Routing & Execution
Orders are:
- Serialized
- Sent over ultra-optimized network paths
- Acked by exchange
Latency here directly affects fill probability.
7. The Latency Budget
Typical breakdown (example):
- Market data parse: ~200 ns
- Strategy decision: ~300 ns
- Risk checks: ~200 ns
- Network send: ~500 ns
Every nanosecond is accounted for.
8. Why Hardware Matters
HFT firms optimize:
- CPU frequency
- Cache hierarchy
- NUMA topology
- NIC placement
Software is written for specific hardware.
9. Determinism Over Throughput
HFT systems prefer:
- Predictable latency
- Consistent execution
Over:
- Peak throughput
- Fancy abstractions
Worst-case latency matters more than average latency.
10. Mental Model
HFT systems are:
Real-time, event-driven, deterministic machines competing on latency
They are closer to embedded systems than web applications.
11. What Comes Next?
Next article dives deep into modern C++ as used in HFT firms:
- Memory
- Performance
- Abstractions without cost
➡ Article 2: Modern C++ for Low-Latency Systems
