A University College London (UCL) and Microsoft-led research team has introduced a new technique for simultaneously synchronizing the clocks of thousands of computers, in less than one nanosecond. The technique, which the team is calling “clock phase caching,” paves the way for the use of all-optical networks by enabling optical switching to take place in data centers.
A heavy and increasing volume of server-to-server traffic currently places high demand on data centers, the extensive networks of connected servers that empower users to perform a wide range of functions online. All-optical networks, using light to transmit and route data to increase data center efficiency, are alternatives to existing technologies that struggle to provide necessary supports.
Previously, however, all-optical network performance and aptitude for use have been limited, given each server’s need to continuously reset its clock time in accordance with all incoming data. By synchronizing the computer clocks of servers connected with optical fiber(s), and using programming hardware to memorize clock phase values, the team eliminated the need for servers to recheck their clock times.
This in turn virtually eliminates the time it takes servers to recover their clocks.
In the past, Moore’s law has allowed cloud providers to accommodate rising demand for network use. Approximately every two years, states Moore’s law, electronic switch integrated circuits double their data transmission with no change to cost and power. Yet the struggle to design smaller, faster, silicon transmitters, coupled with high user demand, continues to call the sustainability and practicality of Moore’s law into question.
“[Our research] has the potential to transform communication between computers in the cloud, making key future technologies like the Internet of Things and artificial intelligence cheaper, faster, and consume less power,” said Kari Clark, a Ph.D. candidate from Optical Networks Group, UCL Electronic & Electrical Engineering, and lead author of the study. Additional members of the research team and contributing authors are from Optical Networks Group, UCL Electronic & Electrical Engineering, and Microsoft Research Cambridge.
The researchers said that reducing computer clock recovery time to less than a nanosecond increases the performance of optical switching compared to today’s state-of-the-art solutions. The time it takes to perform clock phase caching is roughly the amount of time it takes light to travel 30 cm in air.
The study was published in Nature Electronics (www.doi.org/10.1038/s41928-020-0423-y).