The technology, in development for 16 years, will help address the growing demand for better network speed and bandwidth between data centers.
Sixteen years in development, Intel officials are ready to bring silicon photonics to the market.
At the Intel Developer Forum (IDF) this week, Diane Bryant, executive vice president and general manager of the company’s Data Center Group, announced that the chip maker has begun shipping silicon photonics modules, the first products in an expected portfolio that officials have said will help meet the growing demand from cloud service providers for more network bandwidth, better scalability and faster connectivity.
Silicon photonics refers to the use of pulses of light rather than electrical signals over copper wires to move large amounts of data at high speeds and over long distances. Intel’s first two products, the 100G PSM4 (Parallel Single Mode fiber 4-lane) and 100G CWDM4 (Coarse Wavelength Division Multiplexing 4-lane), will target switch-to-switch optical interconnects in data centers and will be available later this year, with adoption ramping next year. The first silicon photonics-based components will enable the transfer of data at 100 gigabits per second over distances of 2 kilometers or more, and will come in a pluggable model and an embedded form factor.
Silicon photonics not only will use light to offer faster connectivity, but Intel’s technology also uses silicon lithography to more precisely align the laser, all of which will lead to greater performance, better efficiency and lower costs, officials said.
The technology will enable organizations to move beyond the scaling and distance limitations of copper wire for transferring data between data centers and within systems, all at a lower cost and greater efficiency, Bryant said during her keynote address at the show.
Intel plans to increase the speed between network switches to 400 gb/s, and officials eventually want to bring optical communications onto the chip, driving at faster connectivity for inside the server.
Intel, which has been shipping the modules since June, is the first step with silicon photonics for helping cloud service providers deal with the rapid growth in data center traffic. According to Bryant, network traffic in the data center is five times that of internet traffic every year, and the amount of traffic is doubling every year.
A change is needed, she said.
“Electrons running over network cables won’t cut it,” Bryant said. “We see a future where silicon photonics optical input-output is everywhere in the data center.”
Microsoft is an early adopter of Intel’s silicon photonics, with officials planning on using it in its Azure cloud data centeres. Kushagra Vaid, general manger of Azure cloud hardare engineering at the software giant, joined Bryant on stage. Vaid said the growth of data traffic in cloud data centers calls for new technologies that can scale better than copper.
“It’s a great way to continue that scaling, to use silicon photonics,” he said.
The plan is to initially to deploy the technology in Azure for switch-to-switch connectivity, Vaid said.
Intel isn’t the only tech company pursuing silicon photonics. IBM engineers a year ago announced they had designed and tested a fully integrated wavelength multiplexed silicon photonics chip, which officials said was a key step to developing 100 gb/s optical transceivers.
IBM’s silicon photonics research is part of a $3 billion effort announced in 2014 to research technologies that will replace traditional silicon chip architectures.
