IP Infusion’s OcNOS 7.0 release is now live for general availability. IP Infusion’s Network Operating System (NOS) running on open networking switches combines the key principles of best-in-class hardware and software. IP Infusion’s Open Compute Network Operation System (OcNOS) provides a rich set of control plane features, providing robust quality, facilitating AI workload transfers, ensuring lower costs and at the same time providing vendors with a best-of-breed selection for open networking hardware platforms. With the OcNOS 7.0 release comes many exciting updates for the data center. Let us explore their highlights and how OcNOS 7.0 further advances the use cases and customer benefits of IP Infusion’s market-leading disaggregated networking solution for data centers.
Defining the Future of AI/ML Fabric Networking
AI infrastructure has entered a new era. As GPU clusters scale to thousands of nodes, network performance is no longer a supporting function — it is the determinant of AI training efficiency, job completion time, and infrastructure ROI. Latency, jitter, and packet loss directly translate into lost productivity at scale.
With OcNOS 7.0, IP Infusion sets a new benchmark for open, disaggregated AI networking.
Built for high-bandwidth GPU fabrics on open switches powered by Broadcom Tomahawk 5, OcNOS 7.0 delivers intelligent congestion control, advanced buffer management, and resilient lossless transport engineered specifically for large-scale AI/ML environments.
More importantly, it brings these capabilities to an open networking model — enabling operators to break free from proprietary stacks while maintaining the ultra-low latency and deterministic performance AI workloads demand.
OcNOS 7.0 combines backend GPU fabric optimization with scalable EVPN-VXLAN multi-tenancy, allowing AI operators to unify performance, isolation, and operational simplicity across their data centers.
As AI infrastructure rapidly evolves, IP Infusion continues to lead with standards-based innovation, empowering cloud providers, AI builders, and enterprises to deploy scalable, high-performance GPU fabrics — without compromise.
Advancing Multi-Site Fabric Leadership with Intelligent EVPN Overlay Extension
Modern data centers are no longer confined to a single site. AI clusters, hybrid cloud architectures, and distributed applications demand seamless connectivity across multiple pods and geographies — without sacrificing scalability, isolation, or operational control.
IP Infusion continues to lead in open networking innovation with intelligent EVPN Layer 3 overlay extension and VXLAN stitching capabilities in OcNOS.
Our architecture enables tenants to extend IP reachability across data center sites while preserving independent control planes and avoiding Layer 2 broadcast domain sprawl. By integrating stitching directly into the fabric’s border leaf nodes, OcNOS eliminates the need for external gateway appliances, simplifying operations and reducing infrastructure cost.
This approach delivers modular scalability, granular tenant control, and efficient inter-site connectivity — all within a standards-based EVPN-VXLAN framework. Operators can selectively extend VRFs and IP prefixes across sites while maintaining strict isolation and optimized tunnel scale.
Raising the Bar in Fabric Visibility and EVPN Policy Control
At scale, operational visibility and policy precision define network leadership. Modern AI and cloud fabrics demand deep, real-time insight into traffic behavior — without adding complexity or external tooling.
IP Infusion continues to lead with integrated visibility and advanced EVPN policy control in OcNOS.
Selective mirroring to CPU with intelligent filtering provides granular, real-time packet insight directly within the fabric. Operators can analyze traffic flows, validate segmentation policies, and troubleshoot overlay or underlay behavior without deploying dedicated capture infrastructure. This built-in observability strengthens operational agility while preserving control-plane efficiency.
At the control-plane level, enhanced EVPN policy capabilities — including Route Target filtering and route-map enforcement — deliver precise multi-tenant isolation and scalable route control. Operators gain fine-grained authority over which routes are imported, advertised, or shared across VNIs, improving security, predictability, and control-plane scale in large VXLAN fabrics.
Together, these capabilities reinforce IP Infusion’s position at the forefront of open, disaggregated networking — delivering the visibility, control, and scalability required for next-generation multi-tenant and AI-driven data centers.
Expanding Open AI Fabric Leadership with Next-Generation ODM Platforms
IP Infusion continues to accelerate innovation in open, disaggregated networking with expanded hardware support in OcNOS 7.0. By adding new ONIE-enabled ODM platforms based on Broadcom Tomahawk 5 silicon, we further strengthen customer choice, supply-chain flexibility, and cost optimization in modern data center deployments.
A key addition is the UfiSpace S9321-64EO, an 800G open networking switch purpose-built for AI/ML fabrics. Delivering 51.2Tbps of switching capacity with 64 high-density OSFP ports, this platform is designed to power next-generation GPU interconnects and large-scale AI clusters.
S9321-64EO front panel view:
By continuously broadening our ODM ecosystem, IP Infusion reinforces its market leadership in disaggregated AI networking — enabling operators to deploy high-performance fabrics with greater agility, vendor flexibility, and long-term infrastructure control.
Detailed information on the OcNOS 7.0 data center features and overall data platforms is available in our Feature Matrix here.
Key Benefits of OcNOS Data Center
- High-performance and scalable data center NOS with Terabits switching bandwidth support
- Deployment proven disaggregated networking solutions
- Open standards-based product, interoperable with existing deployments
- Small footprint resulting from an optimized design
- Simple, all-inclusive licensing scheme
Connect with us to discuss your network needs and see how our carrier-grade disaggregated network operating system delivers on features, scale, and performance at a fraction of the cost compared to traditional OEM solutions.
Additional References
Introducing IP Maestro: Centralized and Simplified Network Management for OcNOS
IP Maestro brings a point-and-click graphical user interface for OcNOS managed networks, providing an intuitive and easy to use interface for complex network management tasks. Informational videos can be found by clicking here.
IP Infusion OcNOS Training Videos, developed and published by online IT training company Orhanergun.net
In an industry first, Orhan Ergun and his staff takes a deep dive into IP Infusion OcNOS. Spanning over 80 lessons, this training course can be found on Youtube, as well as Orhanergun.net. The heart of our training approach lies in a dual focus on a certification of completion and in-depth technology knowledge, equipping learners with not just the “how” but the fundamental “why” behind each concept. Watch the videos by clicking here.
Contact us today to learn more about IP Infusion OcNOS-based networking software.
Alan Huang is the Senior Product Manager, Data Center for IP Infusion.