There are no more obstacles to the distribution of fully optical networks for more efficient data centers. As we continue to rely on the cloud, our need for data centers that can host mission-critical systems has never been greater.
According to a new study conducted by UCL and Microsoft, a new technique that synchronizes computer clocks in less than a billionth of a second can remove one of the obstacles to the distribution of fully optical networks, leading to more efficient data centers.
Data centers, which include tens or hundreds of thousands of connected servers, are the basic technology that allows us to do everything online, from storing movies and photos to serving web pages and online services. However, they face rapidly increasing demand, with server-to-server traffic increasing by 70% each year, which is difficult to meet with existing technologies. Until now, cloud providers have relied on Moore’s law, according to which the complexity of microcircuits and the speed of data transmission doubles roughly every two years at the same cost and power. However, the sustainability of this trend is questioned by the difficulty of making silicon transistors smaller and faster.
Fully optical networks that use light to transmit data are a promising alternative. Until now, their spread has been limited due to the need for each server to continuously adjust its clock time based on the incoming data, which led to overall lower performance.
The study, published on Nature Electronics, shows how the problem can be eliminated through the technique of "clock phase caching", which consists of synchronising the clocks of all connected servers through optical fibers and programming the hardware to store the value marked by the clock so that it does not have to be adjusted.
This discovery, explains Coordinator Kari Clark (Optical Networks Group, UCL Electronic & Electrical Engineering), "has the potential to transform computer communication into cloud, making future key technologies such as Iot and AI less expensive, faster, and less energy-intensive".