Google Principal Network Architect Pushes For Variable-Rate Ethernet
The sultan of search (Google, who else?) wants the Institute of Electronics Inc. (IEEE) to implement a variable-speed Ethernet standard that would make it more efficient to transfer data between data centers and ultimately cut costs. It's a move IEEE has so far been reluctant to take on.
Ron Johnson, a director of product management at Cisco, originally proposed the idea at Light Reading's Ethernet Expo and was disappointed that IEEE chose to stick with the usual single-rate format when it came to 400Gbps. Now it's Bikash Koley, Google's principal architect and manager of network architecture making noise in favor of variable-speed Ethernet.
Koley's reasoning is that Google is unable to use current standards efficiently. The example presented was one of wanting to run a 135Gbps network between data centers. In that scenario, Google would have to create a 200Gbps pipeline, or 150Gbps by link-aggregating 15 10Gbps lines. In either case, it's inefficient and costly. On top of it all, higher speed links usually have to travel shorter distances, which means installing optical repeaters every 600km, Koley says.
According to Light Reading, the technology on the optical side of the equation is already there, as variable-speed transreceivers and flexible-grid ROADMs exist. It's the packet side that's left wanting, in need of a media access control (MAC) layer that can deal with a variable bit-rate physical (PHY) layer.
"We have an adaptive programmable PHY layer ready to go but we don't have the glue to the packet layer," Koley said.
Ron Johnson, a director of product management at Cisco, originally proposed the idea at Light Reading's Ethernet Expo and was disappointed that IEEE chose to stick with the usual single-rate format when it came to 400Gbps. Now it's Bikash Koley, Google's principal architect and manager of network architecture making noise in favor of variable-speed Ethernet.
Koley's reasoning is that Google is unable to use current standards efficiently. The example presented was one of wanting to run a 135Gbps network between data centers. In that scenario, Google would have to create a 200Gbps pipeline, or 150Gbps by link-aggregating 15 10Gbps lines. In either case, it's inefficient and costly. On top of it all, higher speed links usually have to travel shorter distances, which means installing optical repeaters every 600km, Koley says.
According to Light Reading, the technology on the optical side of the equation is already there, as variable-speed transreceivers and flexible-grid ROADMs exist. It's the packet side that's left wanting, in need of a media access control (MAC) layer that can deal with a variable bit-rate physical (PHY) layer.
"We have an adaptive programmable PHY layer ready to go but we don't have the glue to the packet layer," Koley said.
