Arrcus and NVIDIA BlueField-3 DPU Team Up to Boost Enterprise AI Networking Efficiency and Performance

Arrcus, a leading hyperscale networking software company, has announced the integration of its ArcOS® network operating system with NVIDIA’s BlueField-3 Data Processing Unit (DPU) to enhance enterprise datacenters with improved scalability, security, and resource efficiency. This partnership is significant as it addresses the growing demand for AI-driven workloads by optimizing datacenter architecture to handle them more effectively and efficiently. Modern datacenter designs often support multitenancy, where a single physical or virtual infrastructure is shared among multiple tenants. This setup ensures efficient resource sharing, on-demand scalability, and cost reduction. However, the traditional method of executing virtualized network functions (VNFs) like EVPN/VXLAN, routing, IPsec, Access Control Lists (ACLs), and Network Address Translation (NAT) on the host CPU can lead to resource contention and reduced performance. To mitigate this, Arrcus has shifted these functions to NVIDIA's BlueField-3 DPU, which is specifically designed to handle such tasks. This offloading results in increased CPU capacity for mission-critical applications, better performance, and lower power consumption. The integration is powered by NVIDIA’s DOCA software framework, which includes a suite of acceleration libraries. These libraries enable the DPU to process packets at line rate, provide low-latency cloud network connectivity, and manage complex network functions efficiently. The combination of ArcOS and BlueField-3 DPU offers several key benefits: Seamless Connectivity ArcOS, when accelerated with BlueField and DOCA, creates secure overlay tunnels that support robust and scalable multi-cloud networking. Enterprises can seamlessly extend their workloads across different cloud environments, enhancing flexibility and resource utilization. High-Performance Offloading Routing, EVPN/VXLAN, IPsec, ACLs, and NAT functions are now processed on the BlueField-3 DPU, significantly improving throughput and efficiency. This reduces the load on the host CPUs, allowing them to focus on more computationally intensive tasks, thus enhancing overall performance. Optimized Compute Utilization By offloading networking tasks to the DPU, enterprises can free up valuable CPU cycles for other demanding applications. This leads to better compute resource management and more efficient workload distribution, ultimately contributing to higher performance and lower operational costs. Advanced Traffic Engineering & Security Built-in Quality of Service (QoS), telemetry, and security features ensure high performance and reliability while providing robust protection against cyber threats. These functionalities are crucial in maintaining the integrity and availability of the network, especially in high-stakes environments where data security is paramount. Simplified Provisioning and Automation The NVIDIA DOCA Platform Framework (DPF) simplifies the provisioning, configuration, and monitoring of ArcOS on BlueField-3. This streamlined process enhances the agility of cloud-native environments, making it easier for enterprises to adapt to changing demands and deploy new services quickly. According to Shekar Ayyar, Chairman and CEO at Arrcus, the rise of AI-driven workloads necessitates datacenter networking solutions that can deliver high performance, scalability, and energy efficiency. “By leveraging NVIDIA BlueField-3 with our industry-leading ArcOS, we are transforming software-defined networking and unlocking new levels of efficiency for modern datacenters,” he stated. Ash Bhalgat, Senior Director of AI Networking and Security Solutions at NVIDIA, also emphasized the importance of this collaboration. “As enterprises rapidly adopt AI, the demand for scalable, high-performant data center infrastructure has never been greater. ArcOS accelerated with NVIDIA BlueField-3 and DOCA brings new levels of efficiency to enterprise AI factories while supporting zero trust security for advanced AI workloads,” he noted. This integration marks a significant advancement in the field of enterprise AI networking. It not only addresses the challenges associated with managing heavy AI workloads but also sets a benchmark for future developments in datacenter infrastructure. Arcus and NVIDIA are working together to push the boundaries of what is possible, ensuring that enterprises can harness the full potential of AI without compromising on network performance or security. ArcOS on NVIDIA BlueField-3 is currently available for trials and deployment. Interested parties can see a live demo at GTC Paris, Booth S32, or learn more about the solution on the Arrcus website. Industry Evaluation and Company Profile Industry insiders laud the partnership between Arrcus and NVIDIA as a game-changer for enterprise datacenters, particularly those focused on AI. The integration is seen as a critical step towards achieving the high performance and scalability required for AI workloads, all while maintaining robust security measures. Arrcus, headquartered in San Jose, California, has established itself as a leader in disaggregated networking solutions, which allow for flexible and efficient infrastructure management. Backed by top-tier venture capitalists and led by a team of experienced technologists, Arrcus continues to innovate and disrupt the traditional networking landscape. The company’s collaboration with NVIDIA underscores its commitment to delivering cutting-edge, cost-effective solutions that meet the evolving needs of modern enterprises.