5G needs to face diversified service demands, such as industrial control, Internet of Vehicles, VR/AR, smart grid and other ultra-low latency applications, for latency requirement of less than 10ms. The forward plane of such applications must be deployed to the edge DC, providing edge computing to be closer to users and network exports, and relying on the flattening of the network to provide ultra-low latency. Ordinary Web browsing and voice call applications, with a latency of about 100ms, are not sensitive to latency, and can be deployed in regional DCs.
Therefore, the 5G core network realizes complete separation of control plane from user plane, thus bringing about the reconstruction of the entire service network. The core network is no longer a hardware deployment mode in the traditional way of the computer room. Instead, the network and its location can be defined according to software to achieve flexible network function deployment capabilities.
The control plane functions, including mobility management, session management, policy control, and user data management, are all deployed centrally in the central DCs or regional DCs, while the forwarding plane that processes user packets is deployed in the regional DCs and edge DCs. The forwarding plane improves data forwarding efficiency under the unified control of the control plane and improves the end-user experiences. The entire CUPS architecture can flexibly handle different requirements for reliability and latency for different applications.
Different network element functions have different characteristics. In actual deployment, the network elements can be further subdivided. The control plane network elements can be deployed in the central DC or regional DCs as needed. The forwarding network elements, such as the UPF and GW-U, need to be deployed in the regional DCs or edge DCs according to the delay requirements of different service scenarios, which greatly improves the user experiences.
The 5G CUPS architecture constructs a distributed DC based 5GC network. This architecture is completely different from the closed physical architecture of traditional 4G networks, and it is more open, flexible and scalable.
The following figure is a recommended deployment of control/forward separation in a distributed data center. The control plane needs to be centralized to meet the requirements for intensive reduction and cost reduction. The user plane is deployed to the regional DCs or even the edge DCs according to different business scenarios.