Low-Latency Trading System Design

An onsite/super-day system-design round (SWE intern loop): design a trading system given a colocated 'locker box' next to the exchange while your traders work in an office in another city. The discussion centers on the network/infrastructure between the two sites and leans heavily on TCP/IP knowledge. Requirements Design a trading system where you have a locker box (colocated server cage) next to the exchange, while your traders operate from an office in a different city. Address the infrastructure needed to connect the two: network wires/links, bandwidth and latency/speed, and server placement. Expect heavy TCP/IP-level questioning throughout. Notes The round emphasizes low-latency networking trade-offs (colocation at the exchange vs. remote office, link bandwidth vs. latency, where to place compute) rather than a generic web-scale design. Strong fundamentals in TCP/IP and how physical network constraints shape a trading architecture are the main signal; full prompt details beyond the colocation framing haven't been widely circulated. Preparation Review low-latency trading architecture basics: exchange colocation, market-data feeds, order-entry paths, and the latency budget between colocated compute and a remote office. Refresh TCP/IP fundamentals (handshakes, windowing, latency vs. throughput, multicast for market data) so you can go deep when pushed.