In February this year, China Mobile, together with a number of operators around the world, further launched the Open Intelligent Radio Network (O-RAN) Alliance on the basis of the C-RAN concept. The core concept of network access will further promote the cloudification and virtualization of wireless network base stations, and finally realize an open device with decoupling of software and hardware. The NFV virtualization platform needs to further improve performance, optimize energy efficiency, and save costs by introducing accelerated hardware devices, which has formed a consensus in the industry. However, the current accelerator solution has problems such as specific functions and inconsistent interfaces, which cause the software and hardware to be unable to be decoupled. In this context, based on the actual needs of operators, China Mobile proposes a hardware acceleration solution that introduces hardware programmable and unified interface encapsulation. Actively explore NFVI solutions under this concept with Lenovo Group and industry partners, and achieved phased results.

Dr. Yi Zhiling, Chief Scientist of China Mobile Research Institute, said: "China Mobile is committed to advancing the research on a unified hardware accelerator package interface solution, and I would like to thank Lenovo and its partners for their support. This solution is based on ensuring the decoupling of software and hardware. , using programmable general-purpose hardware accelerators to improve the overall energy efficiency of the NFV platform and reduce the overall cost. The wireless cloud platform is the foundation supporting the overall O-RAN solution. We welcome more partners in the industry to join our research. smart radio access network.”

This joint demonstration adopts Lenovo ThinkCloud Accelerator cloud accelerator solution, including Lenovo ThinkCloud NFVI carrier-grade cloud platform software, as well as x86 server and switch hardware; among them, ThinkCloud NFVI can automatically adapt to different manufacturers and different forms of acceleration hardware devices, and through the acceleration The abstraction and virtualization of hardware provides a unified set of API interfaces for network element applications, reduces the complexity of application development and transplantation, and promotes the full decoupling of application software and acceleration hardware.

Acceleration hardware devices in the joint demonstration include Napatech smart network cards based on Xilinx FPGA programmable chips and third-party boards based on ASIC proprietary chips. Xilinx FPGA programmable chips provide flexible hardware logic programming and reconfiguration capabilities, powerful computing performance and energy efficiency, and can be applied to many NFV application scenarios from edge to core. Based on Xilinx FPGA chips, the Napatech smart NIC solution realizes the acceleration of forwarding and computation-intensive operations such as Open vSwitch (OVS) virtual switches and PDCP encryption and decryption of wireless base stations, greatly reducing CPU load, improving performance, and saving costs.

"To meet the agility, scalability and flexibility demands of the rapidly evolving 5G C-RAN, the use of the Xilinx Adaptive Platform on Napatech SmartNICs demonstrates a standard Open API acceleration. The smart NIC solution provides operators with a solution that is easy to manage and deploy through hardware abstraction layer API, and can meet performance requirements in a flexible and economical way. Using the software running on Lenovo servers and cloud platform software The Radisys PDCP protocol stack deploys the accelerator abstraction layer API based on the C-RAN acceleration platform, demonstrating the flexibility, guaranteed latency and performance benefits brought by FPGA-based acceleration in the telecom cloud!"

Jarrod JS Siket, Chief Marketing Officer at Napatech, said: “The current deployment model for NFV for virtualized RAN is to use x86 servers, but it actually hits a major performance bottleneck. Offloading complex networking and security tasks to FPGA-based SmartNICs, Operators can use a common server architecture to maintain NFV mode for VNF processing without compromising the performance, latency costs and power consumption targets required for LTE and 5G applications."

For the cloudification and virtualization of O-RAN wireless network base stations, Radisys provides high-performance and highly scalable 4G and 5G protocol stacks. In this cooperation, Radisys provided a PDCP protocol stack virtualized network element, which supports running in a single virtual machine, and can flexibly expand multiple virtual machines to meet the elastic scaling of processing capacity under different load conditions.

Neeraj Patel, vice president and general manager of software and service solutions at Radisys, said: "Radisys Radisys is very pleased to continue to collaborate with various partners to jointly promote this important demonstration of 5G vRAN for real-world deployment. In this joint demonstration, our high-performance PDCP software protocol suite, including UE simulators and network streamers, was integrated with the software and hardware platforms of our ecological partners, which contributed to the success of this China Mobile RAN acceleration demonstration."