Netronome, a leader in high-performance intelligent networking solutions, today announced collaboration with Red Hat on its Agilio 10/25/40GbE SmartNIC hardware and open software platform to develop future enhancements within the Linux and OpenStack communities. The ongoing development emphasizes open hardware acceleration technologies, aiming to help data center operators better extract more services and enable more users per server rack, a critical requirement in 4G networks and 5G deployments. This includes server racks used for OpenStack, cloud-based NFV and mobile edge computing infrastructures to help deliver more data, video, IoT, augmented reality and vehicle-to-infrastructure services.

The Agilio SmartNIC platform utilizes standard device drivers and data plane acceleration software for Open vSwitch (OVS), which have been developed alongside the Linux kernel community. This brings together the benefits of hardware-based acceleration and the strength of community-developed open source innovations. As a result, network traffic can now be delivered at line rate to virtual machines (VMs) running cloud networking and telecommunication applications while retaining the benefits of software-defined networking (SDN), OpenStack orchestration and VM migration. Based on Netronome benchmark testing, the innovations can deliver more than 25Mpps (million packets per second) while consuming less than a single server CPU core, resulting in up to a 20X gain in efficiency.

An OVS hardware offload solution provides the best of both worlds: Hardware acceleration (fast-path) for high-throughput flows using TC flower offload with SR-IOV, and fall back to OVS kernel datapath (slow-path) for signaling and management traffic requiring advanced Linux kernel features.

The deployment of 5G networks will increase the speed of mobile networks, opening the doors to new services in the areas of IoT, augmented reality and vehicle-to-infrastructure services for autonomous cars. These services demand greater efficiencies from COTS server racks, both in terms of overall throughput and the number of users and applications that can be served per rack. Mobile edge computing seeks to deliver on the latency, quality, and experience demands of these new services. The new hardware-based acceleration and efficiencies of this community collaboration between Netronome and Red Hat can bring benefits to this new paradigm as operators roll out data centers.

"Increasingly, NFV can be viewed as a new standard within the telecommunications industry, enabling communications providers to simplify their network architecture and transform their business models," said Radhesh Balakrishnan, general manager, OpenStack at Red Hat. "While the cost and flexibility benefits of NFV have been publicized, the overhead of virtual switching and SDN overlay networking can lead to performance and scalability limitations in NFV clouds. Coupling a SmartNIC such as an Agilio SmartNIC from Netronome with the flexibility of Linux and OpenStack can help to solve an important problem by enabling increased throughput and reduced server CPU load. We are pleased to have collaborated with Netronome in bringing SmartNIC benefits into the open source ecosystem."

"Netronome is one of the most active participants and contributors in the Linux networking community, especially with mainstream industry-accepted and kernel-based hardware acceleration technologies such as OVS, Linux TC and BPF/XDP," said Sujal Das, chief marketing and strategy officer at Netronome. "We are proud to collaborate with Red Hat in developing the upstreamed SmartNIC acceleration software designed to improve server efficiency and productivity for Telco and cloud operators."

The Agilio SmartNIC family fully and transparently offloads virtual switch and router datapath processing for networking functions such as overlays, security, load balancing and telemetry, enabling servers used for networking and cloud computing to conserve critical CPU cores for application processing while maintaining significantly higher networking throughput.