Aviz Service Node details NVIDIA BlueField-3 DPU traffic offload

Aviz says its Service Node running on NVIDIA BlueField-3 shifts specific control-plane and data-plane packet handling from server CPUs to the DPU to support line-rate traffic steering and subscriber-coherent observability for telco and hyperscale networks.

Research Overview

The blog frames a scaling challenge for telco and large-scale network operators as traffic volume and AI-era workloads increase alongside subscriber-centric analytics. It argues that server-based packet processing for traffic management and observability can become costly and harder to scale.

It presents Aviz Service Node on NVIDIA BlueField-3 DPU as an approach that places traffic optimization and observability functions closer to the network fabric. The goal described is line-rate performance, correlated subscriber visibility, and extensible traffic control.

Key Findings

The blog describes DPU acceleration as a way to offload key traffic management tasks from the server CPU to a programmable acceleration layer. It states this supports scaling observability and traffic optimization without adding CPU cores at the same rate.

It also links the architecture to more accurate visibility by correlating subscriber control traffic with user traffic before directing flows to analytics tools. The blog further describes outcomes as improved probe tool efficiency and reduced infrastructure cost at scale.

Technical Breakdown

According to the blog, Aviz Service Node uses BlueField-3 to process control-plane traffic, accelerate data-plane flows, and steer subscriber traffic to appropriate analytics tools. It describes a combined solution with Aviz network intelligence software and NVIDIA’s programmable DPU platform.

It says control-plane packets such as GTP-C, PFCP, and N11 can be handled on BlueField-3 ARM cores, while GTP-U user-plane flows are steered through the DPU’s P4 programmable datapath at line rate. The blog states that subscriber control and user traffic can then be correlated accurately, directing the right traffic to the right probes instead of sending every packet to every tool.

Operational Impact

The blog lists operational benefits including improved performance through line-rate data-plane acceleration and reduced CPU dependency by handling the heavier traffic functions on the DPU. It also describes balancing probe capacity via subscriber-coherent steering.

For future use cases, it describes programmable extensibility for capabilities such as DPI, AI-based flow policies, and service chaining. It also characterizes the approach as addressing multi-tenant AI-era requirements by maintaining subscriber context while handling data-plane traffic heavy workloads.

Aviz’s blog describes Aviz Service Node with NVIDIA BlueField-3 DPU as moving selected traffic management and observability packet processing closer to the network interface to support line-rate steering and subscriber-coherent visibility while reducing CPU dependency and enabling extensible future capabilities. Blog Signals brief is a fact-based summary of the vendor blog.