Spread Networks, a privately-owned telecommunications provider has launched a dark fiber private network specifically optimized for ultra-low latency for financial industry customers who communicate between Chicago and New York trading centers.
In part, Spread Networks can offer unparalleled levels of latency performance because it has built a brand-new, direct route over the shortest possible route from New York to Chicago, 825 fiber miles long – reducing round-trip latency to 13.33 milliseconds. Up to this point, the lowest latency on the New York-Chicago route was about 15.9 milliseconds, according to Brian Quigley, ADVA Optical Networks senior director.
Where other low-latency connections between those cities uses railroad rights of way, Spread Networks has built along alternate routes, to shave distance, and hence delay. It’s just a guess, but if you want to follow the straightest-possible route between New York and Chicago, you’d follow U.S. Highway 80. That would allow a carrier to relatively easily negotiate rights of way agreements with a few entities and obviously allows easy trenching along the medians.
“Spread Networks has established the competitive standard for trading latency between these two important
financial centers,” said David Barksdale, CEO of Spread Networks (Barksdale was Netscape’s CEO) .
Spread Networks provides customers two strands of dark fiber, which are lit using optoelectronics provided by ADVA Optical Networks. Traffic is kept at layer one to avoid the additional latency if the traffic were carried at a higher level of the protocol stack.
The route terminates at 350 East Cermak Road in Chicago Illinois (telX) and 1400 Federal Blvd in Carteret, New Jersey (Lexent Metro Connect).
As part of the service, ADVA monitors the routes, providing real-time latency reporting. Repeater huts are spaced at 120 kilometers and the route uses low-noise optical amplifiers, dispersion compensation, cut-through switches and no protocol conversion or higher-level switching as part of the effort to achieve the lowest-possible latency performance.
by Gary Kim