Aviz Updates Network Stack with AI

Generative AI (GenAI) is gaining traction in networking, where it can be used to improve automation and management of infrastructure. Aviz Networks has demonstrated the potential for GenAI to improve automation and AI operations (AIOps) with the launch of its Networking 3.0 Stack, which includes its Network Copilot as well as packet broker monitoring capabilities.

This AI-driven upgrade shows the progress that is being made in open networking as well as SONiC (Software for Open Networking in the Cloud) to give customers more flexible, economical solutions to build enterprise networks from the core to the edge.

Aviz’s Networking 3.0 Stack is optimized for open networking environments based on SONiC, a disaggregated network operating system (NOS) that allows customers to run a variety of software components on commercial off-the-shelf (COTS) hardware. The 3.0 upgrade includes Aviz's ONES (Open Networking Enterprise Suite) network operation platform, Open Packet Broker (OPB), and Network Copilot.

Aviz CEO Vishal Shukla says that the company’s software stack will make it easier for enterprises to build, scale, and manage networking infrastructure using SONiC, which has been adopted by large enterprises and webscale providers that want to have more control over their networking infrastructure.

“It enables customers to have choice, control, and cost savings,” said Shukla in an interview with Futuriom.

Maintaining Control Over Data

With Network Copilot, customers can pre-train AI models for their own network benchmarks and allow continuous model training and fine-tuning with real-time data from their networks. Aviz says that this approach will enable use cases such as improving network operations, ensuring compliance, optimizing capacity planning, and more.

One key differentiator of Network Copilot is that customers retain control of the training data, with the option to bring their own data integrations. Network Copilot can integrate with existing data lakes or generate its own data lake while keeping data ownership with customers.

"We will connect the data lake to other applications," said Shukla. "Customers always control the data."

Shukla also pointed out that customers will have the capability to select their own large language models (LLMs) or use an open-source LLM supplied by Aviz.

In its initial release, Network Copilot is available for on-premise and AWS deployments.

Rapid Growth of SONiC Ahead?

Currently, Aviz is working with a handful of very large Fortune 500 companies that see SONiC as having huge potential to reduce the cost and improve the management of their networks, without complex proprietary systems. Aviz has shared some of this information with Futuriom under condition of non-disclosure. In speaking with other potential adopters of SONiC, it’s clear the market has fast growth potential as some enterprise end users tire of the constraints of proprietary NOS lock-in and the expensive operational costs of branded systems.

We expect SONiC to gain real-world consideration from cloud providers, enterprises, and webscale companies. The majority of the hardware vendors have embraced SONiC, making it a viable option for multivendor deployments with developer-friendly networking around it.

And now there’s also a twist: the edge networking boom. With the arrival of AI, there is more activity to deploy new solutions at the edge to handle the influx of the explosion of data generated by AI applications. Expect SONiC systems to play a big role here – and Aviz could benefit.

This recently caught Cisco’s notice. Cisco has become an Aviz investor after spotting the SONiC potential and realizing it needs a hedge against its proprietary NOSs in case open networking takes off.

Founded in 2021, Aviz has backing from prominent investors including Moment Ventures, Accton, Cisco Investments, Wistron, and key angel investors. Aviz is also a member of the Futuriom 50.

Futuriom Take: Aviz has introduced some key AIOps features to its SONiC stack, showing that SONiC can keep pace with the AI-driven network management landscape.