Trading System: Design a Low-Latency Order Matching Engine

Question

## Round 1 - System Design

## Problem

Design a trading system's **order matching engine** that processes buy and sell orders in real time. The engine must maintain an order book and match orders by price-time priority (best bid/ask first; ties broken by arrival time).

Core operations:
- `place_order(order_id, side, price, quantity)` — add a limit order
- `cancel_order(order_id)` — remove an unmatched order
- `get_best_bid() -> (price, quantity)`
- `get_best_ask() -> (price, quantity)`
- `get_trades() -> list[(buy_id, sell_id, price, qty)]` — trades executed since last call

## Example

```
place_order("B1", "buy",  100, 10)
place_order("B2", "buy",  101, 5)
place_order("S1", "sell", 101, 3)
# Match: S1 fills against B2 (best bid=101): trade (B2, S1, 101, 3)
get_best_bid()  -> (101, 2)   # B2 has 2 remaining
get_best_ask()  -> None       # no unmatched sell
```

## Follow-ups

1. What data structures give O(log n) insert/delete and O(1) best-bid/ask lookup?
2. How do you handle partial fills and order amendments?
3. How would you scale this to thousands of symbols with microsecond latency requirements?
4. What consistency guarantees are needed if the engine runs across multiple nodes?

Trading System: Design a Low-Latency Order Matching Engine

Question Details

Full Details

Topics

More from Robinhood