Barefoot Networks: Next-Gen Ethernet Networking
enterprisenetworkingmag

Barefoot Networks: Next-Gen Ethernet Networking

Follow Barefoot Networks on :

Craig Barratt, CEO, Barefoot NetworksCraig Barratt, CEO
Infrastructure in the data center and enterprise went through a fundamental transformation in the last two decades with the rise of virtualization, containers, distributed applications, cloud computing, and much more. The speeds and feeds got faster, yet the techniques and methodologies companies use to build, manage, and debug networks are primarily the same. ASIC, a fixed function switch is no longer considered as a feasible choice for building the networks of the future. A programmable forwarding plane that doesn’t bargain on performance is the key to unlock the revolutionary value of the network and Barefoot Networks is undoubtedly one of the forerunners in the field of programmable networks. Barefoot Networks’ team of visionaries, experienced technologists, and engineers has created a blueprint for designing and operating one of the world’s fastest and most programmable networks.

Barefoot Networks is on a mission to change networking. Their goal is to make programming a network as simple as programming a CPU. With the belief that when the network is fully programmable that is, both the control-plane and data plane are under the control of the end-user, the networking industry will enjoy the same innovative explosion witnessed by the software industry. Barefoot is taking a leap to develop the world’s fastest and most agile networks that will be fully programmable hardware along with compilers, debuggers, and development environment in order to write programs that work. In addition, Barefoot has come up with the world’s fastest switches that are fully programmable for ensuring that the network can also adapt to the emerging needs of the application.

Programmability and performance have been delivered without compromise in other technology domains like computer, graphics, signal processing, and recently in machine learning.

The goal is to make programming a network as simple as programming a CPU


As cloud providers build networks of unprecedented scale, they need to evolve much more rapidly than before, with updated features. To mitigate the scalability issue, Barefoot Networks launched Tofino; the world’s first end-user programmable Ethernet switch. Tofino provides network designers the power of the Protocol-Independent Switch Architecture (PISA). PISA architecture was developed to provide powerful programmability, with its programmable pipeline and recollections taking into account a parcel query table scale that is unparalleled. Rather than purchasing costly switch ASICs with external query table memory, a dominant part of table scale necessities can be met utilizing Tofino.

Barefoot Networks’ unique approach toward networking solutions has enabled AT&T to derive manifold benefits. Leveraging Tofino’s programmability Barefoot Networks solved two significant issues for AT&T. To begin with, time to market and full programmability implied that AT&T controlled its own predetermination recorded as a hard copy of its new data plane, without the risk of relying on a fixed-function silicon vendor to do the data plane programming work. Secondly, programmability provided the foundation on which Inband Network Telemetry could be built to monitor at line-rate any packet header the operator wishes to. By satisfying these two criteria, Barefoot Networks’ Tofino and Inband Network Telemetry played a crucial role in this transformation at AT&T, creating a network infrastructure on which they could offer significant new services.

Barefoot brings the twin pillars of performance and programmability together for the first time in the history of networking. The combination of the Tofino programmable switch chip, the P4 programming language, and Capilano toolset are revolutionary and best in industry performance with the first-ever true switch programmability. To-date, Barefoot has enabled more than 100 new features and applications deployed, in support of improved reliability, visibility, security, artificial intelligence and machine learning workloads, and network storage.