Spread Shaves Ethernet Latency

Low-latency network provider Spread Networks has cut roundtrip delay on its Ethernet wave service between New York and Chicago from 15.75 milliseconds to as low as 14.6 milliseconds, with a service-level agreement of 14.75 milliseconds.

Brennan Carley, senior vice president of product marketing at Spread, says that since the change came into effect earlier this month, “our customers running latency-sensitive strategies tell us that they are getting better quality fills than they were before the latency reduction.”

The vendor could even deliver lower latency, Carley says—although it has no immediate plans for further latency improvements—but technical and commercial issues demand that latency on the wave service remains higher than that of its dark-fiber network, which provides roundtrip latency of 13.33 milliseconds.

“When we came out with our lit wave service, we wanted to be faster than any competitor wavelength but not as fast as the dark fiber, since we wanted it to address the second tier of the market where there is more price sensitivity and less sensitivity to latency,” he says. “But we still believed that we could do better through tuning our dense wavelength division multiplexing equipment in our datacenters and by configuring our network.”

Spread recently completed its network of end-points in different New Jersey datacenters with the extension of the wave network to a facility at 165 Halsey Street in Newark, NJ, and has standardized latency so that clients can achieve the same levels from any of its four end-points in New Jersey. Now that the network is complete, the vendor can begin tweaking components of the network to optimize latency.

Carley says the latency of the wave service will always be higher than its dark fiber because whereas those firms that need the lowest possible latency will typically only use a small number of waves to optimize their network traffic, and will often discard components that make the network easier to manage, Spread has configured its equipment so that each strand of fiber supports a large number of waves, to make the service economical and affordable to as large a client base as possible. In addition, the vendor provides extra services as part of the wave network—such as error detection and correction tools to ensure the reliability required by a large base—all of which add some inherent latency, he adds.