Edgecore and Aviz explain validated SONiC disaggregation for AI NetOps

Edgecore Networks and Aviz Networks describe how ecosystem-driven disaggregation can support AI-ready data center networking, combining validated SONiC builds, orchestration and telemetry, and AI-assisted NetOps in a structured proof-of-concept workflow.

Research Overview

The webinar frames disaggregated networking as a shift from vendor-bundled stacks toward architectures where hardware and software from different sources work together. It also links the approach to the move from CPU-focused infrastructure to GPU-intensive AI workloads and the resulting need for flexible networking validation.

It cites market direction and automation practices, stating that Dell'Oro Group projects disaggregated networking adoption to double year over year. It also references Gartner on automation and AI agents handling runtime networking operations.

Key Findings

Disaggregation is presented as enabling hardware and software independence, cost optimization, lifecycle and supply chain control, and customization at the architecture level. The webinar also attributes adoption momentum to enterprise trends toward automation, observability, and AI-oriented operations.

For open networking deployment, it emphasizes structured validation rather than experimental rollouts, and it positions the Edgecore plus Aviz approach as aimed at production readiness.

Technical Breakdown

The session describes an AIS Certified Community SONiC build validated on Edgecore switches by Aviz Networks. The stated validation steps include feature-level checks, system integration testing, scale and stress testing, long-duration validation, and interoperability testing with third-party vendors.

For orchestration, Aviz One is described as providing YAML-based intent-driven configuration for SONiC fabrics, including BGP configuration, MC-LAG, leaf-spine roles, NTP, IP pools, and automated deployment across Edgecore switches. For observability, streaming telemetry is described as covering CPU and memory utilization, port-level monitoring, ASIC route capacity visibility, IPv4 and IPv6 route tracking, and ACL and fabric utilization metrics.

Operational Impact

The webinar describes a support model labeled “24 by 7,” positioning it as a service approach for open networking environments similar to the Red Hat model for Linux. It also identifies operator-facing capabilities in Network Copilot, including conversational troubleshooting, policy-guarded automation, and correlation across telemetry, configuration, and observability layers.

It further outlines a deep observability stack with an Aviz Packet Broker on Edgecore hardware for traffic aggregation, filtering and header manipulation, tunneling, and packet slicing. It describes service nodes on x86 platforms for packet deduplication, application identification, load balancing, metadata extraction, and packet capture, and it states integration with NDR, IDS, NPM, APM, SIEM, and PCAP tools.

Product and Deployment Model

The webinar presents One Center as an OCP lab that hosts Edgecore hardware in a production-like environment. It describes the ability to run IP Clos or VXLAN proofs of concept, validate feature matrices, test packet broker deployments, evaluate AI NetOps workflows, and access remote GPU-backed Copilot environments.

It also lists validated deployment use cases as IP Clos fabrics, EVPN VXLAN, Layer 2 leaf-spine, edge SONiC deployments, and data center interconnect.

The webinar argues for an ecosystem approach that combines disaggregated networking with validated SONiC, orchestration, observability, and AI-assisted operations through Network Copilot and a structured proof-of-concept lab. Blog Signals brief: this is a fact-based summary of the vendor blog.