SAN FRANCISCO — Microsoft expects to deploy silicon photonics in Azure data centers soon, “initially going for switch-to-switch connectivity,” said Kushagra Vaid, Azure’s general manager of hardware engineering, speaking at the Intel Developer Forum.
Intel began shipping silicon photonics last quarter and celebrated the occasion by mentioning it during an earnings call. The technology got a glitzier launch today from Diane Bryant, executive vice president of Intel’s data center group, during a keynote that featured Vaid as one of a few guest speakers.
Intel’s first commercial silicon photonics products are I/O modules targeting 100-Gb/s connections between data center rows. Eventually, the technology will be used inside servers as well, as those systems begin using 25 Gb/s and 50 Gb/s speeds internally, Bryant said.
“We see a future where silicon photonics optical I/O is everywhere in the data center,” she said.
Intel has been at this game longer than most others, though, nurturing the technology for 16 years, according to Bryant. Along the way, Intel talked about using the technology inside of chips themselves, replacing copper traces with optical connections. But silicon photonics has always had potential in network connectivity as well, and today’s keynote made it clear that Intel wants to participate in that market.
Microsoft has been “making significant investments in Intel FPGAs over the past year,” Vaid said. FPGAs tend to take more space and power than special-purpose chips, but FPGAs can be reprogrammed after installation (“FP” stands for “field-programmable”), making the hardware more fluid.
So, as standards emerge or change, Azure’s team can sometimes reprogram the hardware rather than having to install new chips or boxes. “We don’t have to wait to develop a new generation of hardware,” Vaid said.
Microsoft is also using FPGAs to “accelerate” software-defined networking (SDN) in Azure, Vaid said, although he didn’t elaborate on this point.