Mavenir on Wednesday announced a high performance 5G Core User Plane Function (UPF) accelerated by NVIDIA Mellanox ConnectX-6 Dx SmartNICs. The UPF plays a key role in the 5G network to realize the transformative vision of low latency and high throughput.
Enabling cost-efficient network transformation to 5G, the Mavenir UPF provides a complete set of standalone 5G core user plane functions and combo node support, delivering a unique combination of high performance and capability to support 2G, 3G, 4G and 5G subscribers in a single integrated user plane network function.
Mavenir’s UPF delivers unique value through cloud-native design, container-based architecture, highly optimized packet processing with configurable hardware offloading capabilities, and built-in high availability support. While meeting very high performance demands and low hardware footprint, the Mavenir UPF also delivers on deployment automation and ease of operational maintenance through centralized management system, thus reducing overall total cost of ownership (TCO).
Mavenir has developed a highly optimized UPF solution, both through software and hardware optimization techniques, and has demonstrated performance of up to 524 Gbps using 16 fast path (FP) CPU cores, while reducing overall server footprint by 50% using NVIDIA SmartNIC UPF offloads. NVIDIA ConnectX-6 Dx accelerates the Mavenir UPF through dynamic load balancing, efficient packet processing and intelligent forwarding to keep thousands of customers’ audio, video, and data streams aligned and flowing smoothly at the committed quality of service levels.
Ashok Khuntia, EVP and GM, Packet Computing of Mavenir
We are extremely happy to be partnering with NVIDIA to deliver a high performance and cost-effective UPF to the market for use by both large and small operators around the world.
Amit Krig, Senior Vice President of Software and Network Adapter Products at NVIDIA
Mavenir is leveraging the hardware acceleration of our SmartNICs and DPUs to deliver world-class UPF performance, while freeing CPU cycles to run other revenue-generating applications.
