Barefoot Networks Uses Franz's Allegro CL® to Drive

Artificial Intelligence in Networks

 

Expressive Power of Allegro CL and Common Lisp Enables Machine Learning in Tofino Programmable Chips

 

 

 

Barefoot Networks uses Franz Allegro CL® (Allegro Common Lisp) in the design of the Barefoot Tofino series programmable switch chips to enable Artificial Intelligence, solve complexity, drive superior performance and power efficiency within networks.

 

"At Barefoot Networks we use Lisp (via Allegro CL's Enterprise class development environment) in the design of our Tofino series of programmable switch chips," said Patrick Bosshart, CTO and Co-founder, Barefoot Networks. "Important to us are basic features of the language like its ability to manipulate symbolic data, to represent programs and create embedded languages. We have created embedded languages for building design netlists, and for the gate placement and routing phases of physical design. All of these have the expressive power of lisp seamlessly included. We build custom module generators and layout compilers on a scale significantly larger and more complex than industry standard approaches, which we believe gives us significant competitive advantages in chip area, performance and power efficiency."

 

"The best software programmers in the world choose Common Lisp since it is the most exciting language for deploying Artificial Intelligence solutions. Lisp delivers powerful symbolic computing capabilities and its unique homoiconicity simply makes learning systems more powerful that other programming languages," said Dr. Jans Aasman, CEO of Franz Inc. "We are pleased that Barefoot Networks has tapped into the power of Allegro CL to apply the benefits of Artificial Intelligence to the future of programmable networks."

 

Barefoot Tofino is the world's first end-user programmable Ethernet switch that was built using a P4-Protocol Independent Switch Architecture (PISA). The forwarding logic of Barefoot Tofino resides in the P4 program that the network operator or switch manufacturer loads on the chip, rather than in the silicon. This innovative approach enables Barefoot Tofino switches to employ machine learning to automatically baseline the network performance and detect anomalies. The switch analyzes every packet from every switch and router in the network, in-band at line-rate with an intelligent and flexible triggering mechanism to detect and report events in real time, with nanosecond accuracy.

 

About Barefoot Networks

Barefoot Networks launched in 2016 after two years of developing technology that built switch silicon with a forwarding plane that is defined in software while not compromising on performance. Barefoot empowers network owners and their infrastructure partners to design, optimize, and innovate to meet their specific requirements and gain competitive advantage. In combining the P4 programming language with fast programmable switches, Barefoot has also created an ecosystem for compilers, tools, and P4 programs to make P4 accessible to anybody. Backed by Google Inc., Goldman Sachs Principal Strategic Investments, Alibaba, Tencent, and by premier venture capital firms Sequoia Capital, Lightspeed Venture Partners, and Andreessen Horowitz, Barefoot Networks is headquartered in Silicon Valley. For more information, visit https://barefootnetworks.com/.