Goldman Uses Gear From This Startup to Run Some Data Centers

(Bloomberg) -- Barefoot Networks Inc., a rare semiconductor startup, said a new version of its networking chip will help its effort to shake up how computer networks are built and operated. At least one customer, Goldman Sachs Group Inc., is betting it will succeed.

Barefoot makes switch semiconductors that direct the flow of data between computers. Its Tofino 2 chip is faster and uses less power. Like the original, it can be programmed to do different things via an open-source coding language called P4 that Barefoot helped create.

That flexibility makes Barefoot products different from rival networking chips, such as the Tomahawk range from Broadcom Inc. The startup argues the technology will bring the networking sector up to speed with the innovation sweeping through other areas of computing. Backers include Google and Tencent Holdings Ltd.

Goldman has used some Barefoot-based networking equipment in some of its data centers. The investment bank plans to increase deployments as more mainstream makers of gear, such as Cisco Systems Inc. and Arista Networks Inc., offer switching machinery using Barefoot technology.

“It's an elegant solution. It doesn't solve every problem, and it needs to evolve, but it just keeps improving,” said Josh Matheus, a managing director of technology at Goldman Sachs. “We're pretty happy with it. I see it proliferating.”

For Goldman, there are multiple advantages in using Barefoot-based networking. The ability to customize the chips, and the P4 language, mean new functions can be added quicker, usually within a couple of months, Matheus said. That compares with waiting more than a year for new versions of standard switch semiconductors to be designed and manufactured.

A key goal for computer network owners is real-time monitoring of information as it flows through their data center equipment. Barefoot chips help with this, making it easier to spot and fix outages and slowdowns. Efficiency and security are also enhanced by seeing what happens to information after it enters a network and before it leaves. These extra insights help Goldman make better products such as improved high-frequency trading, Matheus said.

To contact the reporter on this story: Ian King in San Francisco at ianking@bloomberg.net

To contact the editors responsible for this story: Jillian Ward at jward56@bloomberg.net, Alistair Barr, Andrew Pollack

©2018 Bloomberg L.P.