Quick definition Latency is the time it takes for an order to reach the market and for the trader to receive confirmation or a fill. Measured in milliseconds or microseconds, lower latency is faster execution. What it is When you submit an order, your message must travel from your computer to the exchange's servers. The exchange processes the order and sends back confirmation. The round-trip time is latency. High-frequency traders measure latency in microseconds (millionths of a second). Traditional traders measure latency in milliseconds (thousandths of a second). Every microsecond matters in HFT; in other trading, it does not. Why it matters Lower latency gives traders advantages. If you see a price move 1 microsecond before the rest of the market, you can trade on that information. This is why HFT firms invest heavily in latency reduction. Latency also affects execution quality. Large orders might experience latency slippage: the price might move while your order is in transit to the exchange. Latency sources Latency comes from many sources: - Network latency: time for packets to travel - Processing latency: time for your computer and the exchange to process the order - Hardware: some hardware is faster than others - Distance: farther away means longer travel time Practical example You send a market buy order from New York to NASDAQ. The order travels 1,000 miles of fibre optic cable. Latency is approximately 5 milliseconds (the speed of light in fibre is about 200,000 km/s). If you co-locate your servers at NASDAQ (placing your servers at their data centre), latency drops to under 1 millisecond. Co-location and latency HFT firms reduce latency by co-locating servers at exchanges. A server co-located at the NYSE can receive market data and submit orders with less than 100 microseconds of latency. A remote server might have 1-5 milliseconds of latency. See also - Co-location - High-Frequency Trading - Network Infrastructure - Kernel Bypass
