Stock Exchange — Matching Engine

Design a stock exchange. Price-time priority order matching, 100K trades/sec/symbol peak, < 1ms p99 latency, strong consistency.

Time budget: 45 min whiteboard. Draw architecture, estimate numbers, discuss tradeoffs.

Hint 1

In-memory matching per symbol. One process per symbol. Serialized with a journal.

Hint 2

Pre-trade risk holds funds before matching. Trade settler writes double-entry ledger post-fill.

Hint 3

Single-threaded matching > parallel. Determinism + replayable state > throughput.

• +2 Per-symbol single-threaded matching engine with in-memory order book (price-time priority)
• +2 Every mutation journaled (append-only log) before apply; deterministic replay on restart
• +1 Pre-trade risk: balance-hold on order submit; rejected if insufficient
• +1 Trade settler writes double-entry ledger (buyer/seller funds + instrument); idempotent by trade_id
• +1 Market-data fanout via WS/multicast to subscribers; delta updates; < 10ms latency target
• +1 Hot vs cold order book: hot levels around mid-price; deep tail rarely touched
• +1 Circuit breakers on extreme price moves; halt symbol on anomaly
• +1 Clock-drift + jitter: deterministic sequence numbers from matching engine; not wall-clock

Self-score: tally the points you would have mentioned unprompted. 7+ is interview-ready on this problem.

• Matching in a DB transaction (too slow for 100K/sec)
• Parallel matching across symbols with no serialization (order crossings)
• No journaling / replay (state lost on crash)
• Single-machine matching for entire exchange (no per-symbol isolation)
• Skipping pre-trade risk (allows negative-balance order)