“Multi-access edge computing (MEC), as an evolution of cloud computing, has moved application hosting from a centralized data center to the edge of the network, closer to the data generated by consumers and applications, and provides IT and services at the edge of the network close to mobile users. The ability of cloud computing and the use of network capability opening to obtain the advantages of high bandwidth, low latency, and near-end deployment, thereby generating new business and income opportunities, and creating new business models.

“

Multi-access edge computing (MEC), as an evolution of cloud computing, has moved application hosting from a centralized data center to the edge of the network, closer to the data generated by consumers and applications, and provides IT and services at the edge of the network close to mobile users. The ability of cloud computing and the use of network capability opening to obtain the advantages of high bandwidth, low latency, and near-end deployment, thereby generating new business and income opportunities, and creating new business models.

MEC is one of the key technologies to realize 5G low latency and increase bandwidth rate. At the same time, MEC opens the network edge for applications and services, including third-party applications and services, so that communication networks can be transformed into other industries and specific customers Group’s multifunctional service platform.
MEC system architecture

According to the definition of the European Telecommunications Standards Institute (ETSI), the MEC system is divided into two levels: the host level and the system level. The MEC system level network management includes the MEC orchestrator MEO, OSS, and application lifecycle management agent. The host level includes the MEC host and the MEC host. Level network management.

The MEC host is composed of virtual infrastructure VI, MEC platform MEP, and MEC applications. The MEC platform provides an environment for the discovery and use of MEC applications with internal or external services, and through the opening of third-party MEC applications to strengthen the network and business Deep integration.

MEC host-level network management includes MEC platform network management MEPM and virtualized infrastructure network management VIM. The overall architecture of the 5G MEC system is shown in Figure 1.

Figure 1 Overall architecture of 5G MEC system

MEC host deployment plan analysis

With the mature commercialization of 5G and vertical industries, the network needs to access more devices, process massive amounts of data, and meet the needs of low-latency services. The traditional core network centralized deployment model can no longer meet the needs of new services. The network has moved to the edge with the flow of services. It is an industrial trend. The 5G network is originally cloud-based construction, which is lighter and more flexible. Computer room), access office DC, base station computer room as the basic structure of the layered DC computer room layout model has become a common route for the transformation and evolution of the traditional computer room of all operators.

MEC system-level network management needs to coordinate operations between different MEC hosts and between hosts and 5GC (such as host selection, application migration, policy interaction, etc.), which are generally deployed in regional DC (provincial level) or central DC (regional center).

Generally speaking, the deployment location of MEC is mainly for the host-level part of the MEC system. MEC’s ​​ability to support low-latency services and the ability to offload traffic and computing makes it suitable for the three major business scenarios of 5G (enhanced mobile broadband, ultra Reliable low-latency communication and massive large-scale connection of the Internet of Things are all useful. The three major business scenarios, different applications, and different users have different requirements for delay, bandwidth, and computing offload, and the corresponding MEC deployment requirements are also different. All the same.

MEC host deployment should be business-oriented and on-demand deployment, and coordinate with the sinking and distributed deployment of UPF. In actual networking, MEC can be deployed flexibly in accordance with the requirements for operability, performance, or security. From near the base station to different locations on the central data network. But no matter how it is deployed, UPF needs to control the flow of traffic to MEC applications or to the network. Figure 2 outlines the generally available options for the physical location of the MEC.

Figure 2 5G network deployment architecture

MEC is deployed in the access office DC

This mode generally uses a shared computer room between MEC and base station CU, which is deployed behind the base station. Data services are closer to users. The services initiated by the terminal pass through the base station and the MEC host to the Internet/third-party content services. For scenarios where the edge can only meet the demand, the delay can be controlled within 1ms~10ms, such as drone delivery services (10ms, 15Mbit/s), smart venues (10ms, 1Gbit/s), autonomous driving (1ms, 50Mbit/s) Above), remote medical diagnosis (10ms, 50Mbit/s), robot collaboration (1ms, 1~10Mbit/s), remote surgery (1ms~10ms, 300Mbit/s), etc.

MEC is deployed at the edge DC

This mode of MEC is generally deployed in the local network convergence computer room. The logical location is after the UPF/PGW-U (User Plane Gateway), which will increase the delay of a part of the backhaul network, and can provide users with low-latency and high-bandwidth services, such as AR/VR services (20ms, 1Gbit/s), mobile video surveillance (20ms, 50Mbit/s), mobile broadcasting (less than 100ms, 10Mbit/s), public safety (20ms, 10Mbit/s), high-definition video (20ms, 10Mbit/s) /s) etc.

In order to reduce the time delay and enable users to obtain the required content nearby, and to improve the response speed of users visiting the website, MEC equipment usually has CDN (Content Delivery Network) function. Compared with traditional CDN, MEC is closer to wireless Access to the network, the sinking position is deeper. Due to the reduction of physical distance, the latency of natural mobile edge computing is further reduced compared to CDN, and MEC also includes localized computing capabilities and capability opening capabilities, so it has the characteristics of low latency and intelligence, and is used in traditional CDN applications. In addition to scenarios, it will also play a very important role in application scenarios that require intelligence, such as the Internet of Vehicles and smart medical care.

Challenges and solutions in MEC deployment

When deploying MEC hosts, operators generally face the dilemma of weak supporting infrastructure for traditional access office rooms and convergent computer rooms. At the same time, due to the business characteristics and open requirements of edge computing, they face insufficient computer room support and application management/orchestration during the deployment process. Typical issues such as uniformity, security protection, and capability opening.

First, the supporting infrastructure of the computer room is weak. The environment of traditional access office rooms and convergent computer rooms are quite different. If the virtualization infrastructure adopts general hardware under the NFV architecture, most of the computer rooms need to be modified for load-bearing, power supply, air conditioning, etc., and the area of ​​the computer room cannot meet the needs of NFV’s complex edge equipment group. , You can consider adopting more flexible and adaptive enhanced hardware, which occupies a smaller area and has lower power consumption than rack-mounted servers, and is more suitable for edge computer room deployment. And in order to solve the problem of the large proportion of resources occupied by the virtual layer in NFV, the edge adopts the I-layer lightweight solution, supports bare metal containers, and reduces the resource occupation of the management plane.

The second is the lack of uniformity in application management/orchestration. MEC business characteristics determine that there are multiple services such as operator network elements and third-party IT applications in the system, and a unified process and interface need to be adopted for orchestration management. It is recommended that the operator’s cloud management platform unify the management of resources, and flexibly organize and manage them according to different business scenarios: For strong cooperation services, application orchestration and deployment can be performed through a unified portal; for self-management services, application orchestration and deployment can be directly performed at edge nodes.

The third is that the safety protection is not in place. The edge computer room is relatively open and has fewer security protection measures. After the deployment of third-party applications, the telecommunications internal network will face the risk of being attacked. It is necessary to set up a layered isolation and protection plan from the outside, such as an external attack protection plan, a molecular domain isolation plan, Application isolation scheme, etc.

The fourth is to open core network capabilities. In the eMBB business 2C scenario, the third-party APP has the problem of not allowing users to access edge nodes, which needs to be solved by the openness of core network capabilities. When the user’s area cannot be identified through the user message IP address, the auxiliary AF can be used to achieve edge scheduling And on-demand diversion.

The extensive deployment of the MEC platform will bring new operating models to operators, equipment vendors, OTT and third-party companies. In actual deployment, operators can analyze the specific needs of localized business requirements according to the planned business scenarios and the requirements for network delay, security, capacity, etc., according to the scope of services and user characteristics, and choose according to needs. Differentiated deployment schemes require appropriate solutions to the typical problems faced in deployment.