Aviz Service Nodes outline NVIDIA ConnectX support with DPDK

Aviz Service Nodes now support NVIDIA ConnectX NICs using DPDK, expanding beyond x86-only deployments to deliver high-throughput packet processing and AI-ready traffic analytics. The change matters to enterprise and security teams that manage telco, data center, and edge observability at high line rates.

Research Overview

The vendor describes Aviz Service Nodes as software-defined infrastructure for traffic analytics, application identification, and AI-powered insights running on commodity servers. The update adds NVIDIA ConnectX NIC support alongside existing x86 platform compatibility, using DPDK for packet processing acceleration.

The post positions the feature set as applicable to telco cores and 4G/5G traffic correlation, plus data center and edge visibility requirements. It also describes integration with Network Copilot for real-time queries, anomaly detection, and trend analysis.

Key Findings

With ConnectX support, Aviz Service Nodes claim throughput of 100 GbE and above via hardware-accelerated packet processing. The capabilities continue to include GTP correlation, subscriber-aware metadata, application identification across more than two thousand applications, and packet deduplication.

The post also highlights a ConnectX-specific feature: hardware-accelerated VXLAN header stripping. It states this capability is not available on Intel E810 NICs, and ties performance claims to near-zero CPU usage and sub-microsecond latency.

Technical Breakdown

Aviz Service Nodes use DPDK with NVIDIA ConnectX by initializing DPDK poll-mode drivers to access ConnectX high-speed queues. The workflow captures and processes traffic at speeds exceeding 100 GbE and routes it into deep packet inspection, GTP load balancing, and flow distribution.

For VXLAN, the vendor states ConnectX offloads the removal of 50-byte VXLAN headers that include Ethernet, IP, UDP, and VXLAN. It says configuration is handled through the ASN user interface, and provides performance comparisons versus software-based stripping on x86 cores.

Operational Impact

The post describes deployments on commodity servers from Dell, Cisco, SuperMicro, and Wistron. It claims cost and architecture flexibility by eliminating proprietary appliances, stating more than 50 percent TCO savings, and enabling hardware upgrades without rebuilding the observability stack.

It also describes how Network Copilot consumes processed traffic insights from Aviz Service Nodes for real-time queries, anomaly detection, and trend analysis. The vendor connects ConnectX acceleration to GTP-aware load balancing and subscriber correlation at scale for 4G and 5G traffic processing.

Leadership Perspective

The vendor frames the update as maintaining a hardware-agnostic, modular design that supports multiple commodity server vendors and now NVIDIA ConnectX NICs. It states the approach avoids vendor-specific dependencies by keeping capabilities tied to the service node design rather than a single hardware platform.

In the release narrative, the combination of ConnectX packet acceleration and Network Copilot AI insights is presented as the basis for faster traffic processing and interactive analysis. The post describes the result as network visibility suitable for telco, data center, and edge environments.

This update adds NVIDIA ConnectX NIC support to Aviz Service Nodes using DPDK, including claims of 100 GbE+ throughput, GTP correlation, and hardware VXLAN header stripping, plus integration with Network Copilot for AI-driven analytics. For enterprise decision-makers, it outlines a commodity-server deployment model intended to reduce reliance on proprietary appliances while preserving traffic intelligence functions. Blog Signals brief is a fact-based summary of the vendor blog.