Recently, on the 5G summit, Ms. Yang Rui, technical director of Cloud&Core Network Product from ZTE, pointed out that only to combine software and hardware together, work on 360° enhancement, can meet the extreme experience requirements of 5G services.
The key performance indices of 5G network are more diversified. The user data rate reaches Gbps, the peak rate reaches tens of Gbps, and the latency low to 1 ms. The user connection data density reaches 1 million connections/km², and the mobile speed should reach 500 km/h. The 5G network adopts distributed architecture so the user plane can be as close to the end users as possible to achieve large bandwidth and lower latency. The evolution of the telecom cloud to a distributed cloud is an inevitable trend of 5G network development. ZTE focuses on R&D innovation in hardware customization, cloud and service acceleration to help operators build the new-generation cloud network infrastructure with low latency, high bandwidth and high performance.
In terms of hardware, ZTE fully utilizes latest hardware products from industry partners, and also produces self-designed chips and servers, to provide ubiquitous computing power with maximum performance. The computing power, memory, IO and heterogeneous expansion capabilities of the new generation G4X servers are greatly improved. Different hard disk back planes and extended suites are used in a way of building blocks to flexibly match different application scenarios, such as database, NFV and AI. Yang Rui stressed that for the edge UPF, ZTE also provided the U9400 forwarding server, with a built-in switching module, to support 420Gbps forwarding performance. In the chip field, ZTE self-designs chips for specific service scenarios, with data-stream-driven architecture, the energy-efficiency ratio is 20 times that of GPU. It supports variable core expansion and can be customized, such as DPU chips, AI acceleration chips, and video acceleration chips.
In terms of cloud platform, ZTE provides TCF cloud base, which uses DPDK, SRIOV, NUMA binding, OS optimization and huge page technologies for all-round telecom enhancement. OpenStack and Kubernetes are deeply integrated to save management resources, improve resources utilization and provide users with unified management interface and user experience. ZTE has always been committed to technical innovations. NEO cloud cards are released this year, which offload the management modules and virtualization layer of the cloud platform from the server to the cloud card, simplifying cloud platform deployment and saving server CPU consumption. In addition, the NEO card and the server are controlled unidirectionally through the PICe interface to achieve physical isolation and improve the system security level.
In terms of service, ZTE enhances the data forwarding capability in terms of software and hardware to meet the requirements of 5G eMBB and URLLC. With the software-only acceleration solution of UPF, a single server can provide the 60Gbps processing capability with a delay less than 100 µs. For industrial control, autonomous driving and other ultra-low latency applications, ZTE’s UPF, which is accelerated based on the FPGA smart NIC, greatly reduces the bottlenecks of CPU computing, memory reading, and PCIe bus, thus improves the performance density of a single server. Compared with the UPF for software-only acceleration and the UPF for smart NIC acceleration, single server throughput is increased to three times, the latency is reduced by 90%, the power consumption per Gigabit is reduced by 55%, and the performance is industry-leading.
Finally, Ms. Yang Rui said that 6G networks are not far away from us, the technical revolution for 5G is only the beginning. ZTE will introduce intelligence and automation to achieve the ODICT integration of operation technology, data technology, IT and communication technology.