- This Samsung phone is the model most people should buy (and it's not a flagship)
- The 50+ best Black Friday Walmart deals 2024: Early sales live now
- How to Dockerize WordPress | Docker
- The smartwatch with the best battery life I've tested is also one of the cheapest
- One of the most immersive portable speakers I've tested is not made by Sony or Bose
Aviz gives its enterprise SONiC offering an AI boost
Aviz Networks continues to add features to its enterprise-grade SONiC offering as it works to make the open-source network operating system a more feasible option for customers.
The latest addition is an AI-based package called Network Copilot that’s aimed at improving network operations, management and capacity planning. The ultimate goal is to help network professionals integrate AI into their daily jobs and streamline the way organizations operate their future networks, according to Vishal Shukla, CEO for Aviz.
“Right now, network operations are template driven. But going forward, operations will be data driven, and AI will help you to manage your network from the observability and the orchestration point of view, ultimately making networks open and much easier to manage,” Shukla said.
Network Copilot will work by gathering data from Aviz’s core Networking 3.0 Stack package, which includes support for intent-based networking technology, network telemetry, anomaly detection and the vendor’s Open Networking Enterprise Suite (ONES). The ONES suite of management software includes SONiC configuration templates for data-center leaf/spine configurations and supports standard networking technologies such as EVPN, VxLAN, and border gateway protocol (BGP).
ONES compiles a network’s hardware and software inventory by gathering telemetry from switches in the network. It supports multi-vendor NOSes such as Nvidia Cumulus Linux, Arista EOS, and Cisco NX-OS, as well as switches that utilize standard OpenConfig telemetry, according to Aviz.
From the Networking 3.0 Stack information, Network Copilot customers can pre-train their own large language models (LLM) for network benchmarks for functions such as network operations, compliance, capacity planning, bandwidth management and troubleshooting. Customers then can fine-tune those models with real-time data from their networks, Shukla said.