AI-Ready Network Observability

Complete Control, Half the Cost, Built to Last! Data Insights and Tool Compatibility with Vendor-Agnostic Packet Broker and Service Nodes

February 24, 2025

About Aviz Networks

On a mission to deliver networks for AI and AI for networks, Aviz Networks was founded to modernize and transform the networking software stack, addressing the evolving demands of data centers, edge, and GPU networks as they scale and integrate AI. Enterprises rely on Aviz for deploying vendor-agnostic Community SONiC solutions, enhancing network observability, and introducing AI-driven management capabilities. These solutions empower customers with hardware choices, operational control, and substantial cost savings. Aviz's innovations have been successfully deployed by leading e-commerce platforms, telecommunications providers, retailers, GPU-as-a-service providers, cloud service providers, and diverse enterprise networks.

Why Now is the Right Time to Consider a New Network Observability Solution?

"The networking landscape is rapidly evolving, driven by commoditization, open standards, and the rise of AI-driven operations. Adopting an open approach to Network Observability empowers businesses with a holistic, AI-driven solution that integrates Network Copilot™, Open Packet Broker (APB), and Aviz Service Nodes. This unified approach enables businesses to capture, process, and analyze network traffic in real time, providing deep observability into network behavior, optimizing operations, and enhancing security. OPB efficiently aggregates and distributes traffic, Service Nodes perform advanced packet processing and analytics, and Network Copilot leverages AI to deliver actionable insights—all working together to provide full-stack observability."

Benefits you get

01

Enhanced Network observability with AI: Our Network Copilot elevates observability across your networks, integrating AI to analyze and manage network performance more effectively. This ensures that you can not only monitor but also analyze network behavior, optimizing operations and enhancing security.

02

Choices: Select hardware based on your needs, or better reuse your old hardware. No more putting all your eggs in one basket of hardware vendors.

03

Control: Break free from single-vendor lock-ins with our flexible software-driven solutions, which are agnostic of hardware. This software-driven approach makes your infrastructure robust and future-proof, giving you the freedom to choose without the hassle of overpaying for hardware refresh even when you are not adding more functionalities.

04

Cost Savings: Reduce network observability costs by leveraging an open, vendor-neutral approach. Avoid expensive proprietary solutions and extend the lifecycle of existing hardware while optimizing network monitoring expenses with AI-driven efficiencies.

Technical Overview

The Unified Fabric Manager enables AI-driven network observability by capturing TAP & SPAN traffic from data centers and 4G/5G networks, processing it through Open Packet Brokers (OPB) for aggregation and distribution, and analyzing it via Service Nodes for packet deduplication, DPI, application identification, GTP correlation, and KPI analysis. The extracted metadata is then fed into Network Copilot and customer tools for real-time insights, troubleshooting, and optimization. This scalable solution enhances network performance, reduces operational costs, and automates observability, making it ideal for modern AI-driven network operations.

Open Packet Broker

Aviz's Open Packet Broker (APB) is the core of its Network observability Stack, designed to future-proof network monitoring through disaggregation, which separates the data plane from the control plane for greater flexibility and scalability. This approach reduces costs by enabling upgrades without a complete system overhaul.

Filtering

Filtering traffic based on select criteria (L2-L7) of interest to the tool needs

Packet Labeling

Track packets on ingress to identify source and mark packets at egress for tools

Aggregation & Replication

Combine data from multiple sources towards single/multiple tool

Packet Slicing

Discard secure or excess payload thereby reducing the disk storage utilization.

Load Balancing

Symmetric Load balancing based on IP (v4/v6) header

Tunneling

VXLAN Header stripping within and Encap/Decap across DCs

Aviz Service Nodes

The Aviz Service Node is a comprehensive network observability and analytics platform designed to enhance real-time packet processing, optimization, and monitoring across Telco and datacenter environments. It supports packet decoding, deduplication, and modifications, ensuring efficient traffic handling by removing duplicate packets, adding timestamps, and encapsulating headers at wire speed. The platform employs symmetric inner and outer header-based load balancing to maintain session consistency for analytics and performance measurement tools. With metadata export capabilities, it integrates seamlessly with external tools via Kafka and ELK stack, enabling detailed analysis through JSON-formatted packet metadata. It also provides advanced KPI calculations, tracking bandwidth, latency, and retransmit counts to assess network health.

Aviz Service Node Capabilities

Advanced packet processing for Telco and datacenter networks

Deep Packet Inspection

Filtering traffic based on select criteria (L2-L7) of interest to the tool needs

KPIs

Track packets on ingress to identify source and mark packets at egress for tools

Packet Deduplication

Combine data from multiple sources towards single/multiple tool

Application Identification

Discard secure or excess payload thereby reducing the disk storage utilization

Mobile Core

Line rate DPI and GTP Correlation for Telco networks supporting 4G-LTE, 5G-SA, 5G-NSA

Tunneling

VXLAN Header stripping within and Encap/Decap across DCs

Customer Case Study

"Deep Network Observability Improves 5G Mobile Subscriber User Experience - add link and image (KDDI Paper)"

Connect with our experts to explore how Aviz can simplify and scale your AI infrastructure.