Colocation and Network Topology

Signals travel through fiber-optic cable at roughly two-thirds the speed of light, which sounds fast until you realize that distance still costs measurable time at the microsecond scale competitive trading operates at — a round trip to a data center 100km away costs roughly a millisecond, an eternity compared to intra-data-center latencies measured in nanoseconds to low microseconds. Colocation means renting server rack space physically inside (or immediately adjacent to) an exchange's own data center, so a firm's trading server is as close as physically possible to the exchange's matching engine, eliminating the long-haul network delay entirely.

But colocation alone isn't enough — network topology, the specific arrangement of switches, cables and cross-connects between a firm's servers and the exchange gateway, determines whether that proximity actually translates into an advantage. Two firms in the same building can have meaningfully different latencies if one's cable run and switch hops are shorter, or if one pays for a direct cross-connect while the other routes through a shared switch with extra hops (each hop adding processing delay). Exchanges typically try to equalize this by offering identical-length cables to every colocated client, precisely because topology, not just the building, decides the race.

Concrete illustration. Two firms both colocate in the same exchange data center. Firm A pays for a direct fiber cross-connect straight to the exchange's matching-engine cabinet; firm B connects through the data center's shared switch fabric, adding two extra network hops. Even though both are "in the building," firm A's round-trip order latency can be several microseconds faster — often decisive in a market where the fastest quote wins the trade.