It turns out Facebook is into networking that is more than just social, unveiling an open-source, software
-defined networking switch this week at the Gigaom Structure Conference in San Francisco. Some industry-watchers are speculating that Facebook's new switch could ultimately disrupt the networking equipment market, one of the last dominions of proprietary hardware. If it succeeds, the Facebook switch could even pose a threat to network-equipment market-leaders including Cisco
and Juniper Networks.
The switch, code-named Wedge, is being tested in Facebook's data centers. The company plans to have it in full production mode and released to the Open Compute Project (OCP) by the end of the year, said Jay Parikh, Facebook's vice president of network engineering.
Facebook founded OCP in 2011 to promote the use of open-source data center hardware. By developing open-source server designs, Parikh said, the company has saved $1.2 billion in infrastructure costs.
Wedge Overcomes Proprietary Limitations
With the proprietary networking gear, Parikh said, “we were running into a lot of things that slow down. It doesn’t give us the flexibility we want -- the control we want -- the ability to configure things, monitor things, adjust [network] flows with things like Messenger, Instagram, Search, and all the other workloads we have. Being able to manage all that on the network is a very rich area for optimization.”
The Wedge is a "top of rack" switch, built into each rack of servers at a Facebook data center. It has 16 40-Gigabit Ethernet ports and can be expanded to 32 ports, Parikh said. Facebook built its own chassis -- painted Facebook blue -- that is optimized for cooling and can have dual AC or DC power supplies.
The Wedge switch is built using the same microprocessor platform that is in Facebook's servers, and the company has developed its own network operating system, FBoss, which is based on the same Linux variant used in its other servers.
"It is no longer a switch. It’s just a server,” Parikh said of Wedge at the Structure event. With all the shared hardware and a customizable operating system similar to that on the other data center servers, costs are reduced and operations are simplified, he said.
Control and Flexibility
In a post on Facebook's Engineering Blog, Yuval Bachar and Adam Simpkins extolled the virtues of using FBoss on the Wedge.
"The service layer in 'FBoss' allows us to implement a hybrid of distributed and centralized control. We ultimately want the flexibility to optimize where the control logic resides, which in turn will allow us to get higher utilization on our links, troubleshoot easier, recover from failure faster, and respond more quickly to sudden changes in global traffic," they said in the blog post.
"For example, using a central controller, we can now find the optimal network path for data at the edges of the network. By doing so, we've managed to boost the utilization of edge network resources to more than 90 percent, while serving Facebook traffic without a backlog of packets."
Because FBoss is customized just for Facebook, the company will mainly open-source the Wedge hardware design and just some of the core FBoss components for control, the company said. Several software-defined networking start-ups have already expressed interest in implementing the design into a complete off-the-shelf package.
Cisco, for its part, said it wasn't worried by open-source projects like Facebook's Wedge.
"That’s because the approach is loaded with hidden hard and operational costs," a Cisco spokesperson told Barron's in an e-mail. "For example, networking capital equipment outlays typically constitute only 30 percent of the cost of running networks. Labor costs constitute 50 percent, and will increase with the white box approach."