VelOS-DC

VelOS DC is a data centre network operating system built from the ground up on Linux -- not layered on top of it. It integrates natively with compute environments, automation frameworks, and AI/GPU cluster fabrics, delivering hyperscaler-grade networking on whitebox hardware -- with seamless integration into existing infrastructure through open standards.

Download Brochure

Why VelOS-DC

Data centre networks are not general-purpose infrastructure -- they are the foundation on which every compute workload, every storage operation, and every AI training job depends. The network operating system that runs the fabric must be as native to the data centre environment as the servers it connects. VelOS DC is Linux -- not a proprietary system with a Linux compatibility layer bolted on. The kernel is the source of truth for network state, and every tool a Linux engineer already knows works out of the box. Ansible playbooks configure switches the same way they configure servers. Systemd manages network services the same way it manages compute services. Shell scripts, cron jobs, and standard Linux utilities operate without translation layers or proprietary CLI abstractions. This is what Linux-native means: the network operates as a first-class member of the compute environment, not an isolated appliance managed through a separate toolchain. VelOS DC runs on NeoX data centre switches built on industry-standard Broadcom silicon -- from Trident 3 leaf switches to Tomahawk 4 hyperscale spines. Operators choose their hardware, retain full flexibility, and avoid the proprietary dependencies that legacy vendors impose. The result is a data centre fabric that delivers hyperscaler-grade capabilities with the operational simplicity of standard Linux, the freedom of open hardware, and seamless integration into existing compute and automation environments.

Request A Demo

Key benefits

What Makes VelOS-DC Powerful

VelOS-DC is a Linux-native network operating system built for modern data centres, where performance and reliability are critical. It enables seamless use of standard tools like Ansible and systemd, simplifying automation. Running on NeoX switches with Broadcom silicon, it offers hardware flexibility, scalability, and high-performance networking without vendor lock-in.

Scalable Fabric Architecture

VelOS DC supports leaf-spine fabric architectures that scale from small data centres to hyperscaler-grade deployments. The hardware ecosystem spans Broadcom Trident 3 silicon for cost-effective leaf switches through to Tomahawk 4 silicon for ultra-high-bandwidth spine switches, enabling operators to build fabrics that grow with their infrastructure without changing their NOS.

Compute-Native Integration

Linux-native means every tool a Linux engineer knows -- Ansible, Puppet, systemd, cron, shell scripts -- works natively on VelOS DC. There is no proprietary CLI translation layer between intent and execution. Network switches operate as standard Linux hosts, integrating seamlessly into compute-centric operational workflows and automation pipelines.

Open Architecture, Zero Lock-In

VelOS DC runs on NeoX DC switches and other whitebox platforms built on industry-standard Broadcom silicon. Standard protocols, standard interfaces, and standard automation frameworks mean operators retain full hardware choice and avoid proprietary dependencies that limit future flexibility. Choose your hardware, keep your options open.

AI-Ready Performance

RDMA over Converged Ethernet version 2 (RoCEv2) and lossless Ethernet are production features shipping today -- not roadmap items. GPU clusters and AI training workloads get the low-latency, lossless fabric they need now, on open hardware, with full integration into existing data centre automation and monitoring workflows.

VelOS DC Architecture

The VelOS DC architecture diagram illustrates the separation of control plane, data plane, and management plane within a Linux-native network operating system. The Linux kernel serves as the foundation, with a hardware abstraction layer providing uniform access to Broadcom switching silicon. The protocol stack, automation interfaces, and streaming telemetry services operate as standard Linux processes, enabling native integration with compute-environment toolchains.

LINUX-NATIVE NOS

Linux kernel as source of truth

Network state lives in the kernel, not in a proprietary database. Standard Linux networking commands (ip, bridge, tc) interact with the network directly.

Hardware abstraction layer

Supports Broadcom Trident 3 through Tomahawk 4 switching silicon, providing a consistent operational experience across the full range of NeoX DC switches.

Native Linux tool integration

Standard Linux networking tools work out of the box. No proprietary CLI required for day-to-day operations.

Full protocol stack

Border Gateway Protocol (BGP), Open Shortest Path First (OSPF), and Ethernet Virtual Private Network (EVPN) with Virtual Extensible LAN (VXLAN) overlay for scalable data centre fabrics.

Streaming telemetry

Real-time network state exported via gRPC Network Management Interface (gNMI) for integration with modern monitoring and observability platforms.

Programmable data plane

Hardware-accelerated forwarding with support for custom filtering, Quality of Service (QoS), and Access Control List (ACL) policies.

Request A Demo

AI NETWORKING CAPABILITIES

Built for AI/GPU Cluster Fabrics

he following capabilities are production features shipping today on VelOS DC -- not roadmap items. Operators deploying GPU clusters and AI training infrastructure can qualify VelOS DC for their lossless fabric requirements now.

Request A Demo

RDMA over Converged Ethernet v2 (RoCEv2)

RoCEv2 enables GPU-to-GPU communication with minimal latency and near-zero CPU overhead. In distributed AI training environments, thousands of GPUs must synchronise gradients across the fabric after every training iteration. RoCEv2 bypasses the traditional TCP/IP stack, delivering the consistent low-latency performance that large-scale AI workloads demand.

Lossless Ethernet (Priority Flow Control / Explicit Congestion Notification)

Priority Flow Control (PFC) prevents packet drops on priority traffic classes by sending pause frames to upstream devices when buffer thresholds are reached. Explicit Congestion Notification (ECN) signals congestion to endpoints before drops occur, enabling proactive rate adjustment. Together, PFC and ECN create the lossless Ethernet fabric that RDMA workloads require -- ensuring that every packet reaches its destination without retransmission.

Data Centre Quantised Congestion Notification (DCQCN)

DCQCN is an end-to-end congestion control protocol that combines ECN signalling with fine-grained rate adjustment at the source. It prevents congestion hotspots in large-scale AI training clusters by dynamically throttling senders before queue depths trigger PFC pause frames. DCQCN is essential for maintaining consistent fabric performance across hundreds or thousands of GPU nodes operating in parallel.

PROTOCOL AND FEATURE DEPTH

VelOS DC supports a comprehensive set of data centre networking protocols, automation interfaces, and monitoring tools.

Category	Protocols and Standards
Category Layer 2	Protocols and Standards Spanning Tree Protocol (STP), Rapid STP (RSTP), Multiple STP (MSTP), Link Layer Discovery Protocol (LLDP), Link Aggregation (LAG/LACP), Virtual LAN (VLAN)
Category Layer 3	Protocols and Standards Border Gateway Protocol (BGP), Open Shortest Path First (OSPF), static routes, Equal-Cost Multi-Path (ECMP), Virtual Routing and Forwarding (VRF)
Category Overlay	Protocols and Standards Virtual Extensible LAN (VXLAN), Ethernet Virtual Private Network (EVPN)
Category Automation	Protocols and Standards Ansible, Terraform, Python, REST API, gRPC Network Management Interface (gNMI), Network Configuration Protocol (NETCONF
Category Monitoring	Protocols and Standards sFlow, Simple Network Management Protocol (SNMP), syslog, gNMI streaming telemetry

What VelOS-DC Brings to Your Network

Product Type	Ports Supported	Chipset
NeoX NS7326-56X	48x 25G SFP28 + 8x 100G QSFP28	Broadcom Trident 3	Download pdf
NeoX NS7726-32X	32x 100G QSFP28 + 2x 10G SFP+	Broadcom Trident 3
NeoX NS9736-64D	64x 400G QSFP56-DD	Broadcom Tomahawk 4
NeoX NS9716-32D	32x 400G QSFP-DD	Broadcom Tomahawk 3