MEC: Past, Present and Future

 

 

Since the appearance of the first computer in 1946, information technology has been upgraded in a rapid, iterative fashion. In the past 10 years, 4G network and cloud computing were developed. 5G will deliver high-quality services such as eMBB, URLLC and mMTC. Meanwhile, 5G faces the following new challenges:
—In the eMBB scenario, the 4K/8K content brings great pressure to transmission/bearer network.
—In the mMTC scenario, the massive data generated by massive connections leads to the requirement for distributed computing power.
—In the URLLC scenario, the real-time computing power of the cloud is insufficient to meet the ultra-low latency service requirement down to milliseconds.
—Digitalization and AI (e.g. image recognition, machine decision-making, and AR/VR) create the need for more and more computing power, resulting in increased terminal costs.      
—For the sake of data security and privacy, some industrial applications require local data processing and storage.

To cope with the above challenges, the industry has focused its attention on MEC. MEC brings computing power to the edge of the telecom network to localize services, which effectively reduces service latency, bandwidth overhead, and terminal costs, improving service experience and data security, and providing effective support for new human-centered services and object-centered internet of everything (IoE) applications.

The Rise of MEC in 4G Era  

MEC is not a new product of 5G. In the 4G era, many mainstream operators launched MEC pilots to explore its technologies and business models. The MEC pilot projects added computing, storage, and processing capabilities at the edge of a mobile network to carry different industrial applications, such as CDN, video surveillance, and face recognition. These attempts have laid a foundation for the vigorous development of MEC in the 5G era. However, the MEC solution in the 4G era had some technical weaknesses:
—The MEC standards were not complete. The MEC reference architecture, application lifecycle management and O&M framework were defined by ETS in the 4G era, but there was a lack of multi-access edge computing system, support for network slicing, interface specifications and orchestration & management.
—Since the 4G core network doesn't apply control and user plane separation, the edge-side offload and interconnection solution is complicated. The traffic offload solution at the wireless side is lacking in terms of supervision, security and charging, making it difficult to be commercialized.  

In addition, the ecosystem of the MEC applications is not complete and enterprise users' enthusiasm in participation is rather low; therefore, MEC has not been widely used in the 4G era. 

MEC: Best Partner of 5G

Compared with 4G, 5G increases connection density and traffic density by dozens or even hundreds of times, reduces the latency to several milliseconds, and greatly improves network quality. This allows upgrades of existing 4G services, improvement of digital experiences and introduction of new services. ETSI defines seven MEC application scenarios, including intelligent video acceleration, video stream analysis, augmented reality (AR), assistance for intensive computation, enterprise, connected vehicles, and IoT gateway. In addition, it defines in details the MEC reference architecture (Fig. 1), end-to-end edge application mobility, network slicing support, lawful interception, container-based application deployment, V2X , Wi-Fi and fixed network capability exposure, better supporting the implementation of MEC at the standard level.

 

From the perspective of major MEC application scenarios, the major issues to be solved are connection and local traffic breakout, network QoS guarantee, computing power supply, capability exposure and sharing, and rapid application deployment. Therefore, the key to MEC is to build flexible networks, various clouds, and abundant capabilities, and, on this basis, create an application ecosystem to provide services for customers.
—Flexible network: Full connection, edge offload and quality assurance
    MEC supports various network connections, including 4G, 5G, Wi-Fi, fixed network and NB-IoT, meeting the access requirements of various terminals. It allows traffic breakout at the edge, implements traffic localization, reduces the pressure on the bearer/transmission network, solves enterprises’ networking problems and eliminates their concerns about data security. 5G network supports flexible deployment of UPFs in different locations and provides multiple offloading policies. Network quality assurance is provided on demand, and the service quality requirements in terms of bandwidth and latency are guaranteed through the QoS and/or network slicing mechanism.
—Diverse clouds: Support for diversified hardware, on-demand provision of various capabilities, and rapid application deployment
    MEC supports various computing resources and acceleration hardware (smart NIC, GPU, and FPGA). It provides computing power as required. MEC provides bare-metal servers, VMs and containers as well as various acceleration resources to accelerate networks, image/video analysis, and encryption and decryption. Based on the characteristics of edge equipment rooms, MEC uses a lightweight deployment solution, and supports one-click provisioning and unattended O&M.  
—Rich capabilities: Exposed network capabilities and simplified application development
    By exposing network information and capabilities, such as TCPO, radio network information service (RNIS), QoS capability, bandwidth capability, and location capability, MEC allows services to be aware of the network so as to enhance service experience. General IT services are abstracted for application developers to simplify and accelerate application development, such as video recognition, low-latency video, IoT device management, video optimization, and app registration management.
—Building an application ecosystem to meet the needs of individual users and industry users
    5G will redefine a wide variety of industries. Different industries face different problems in the informatization and intelligentization process. Operators and equipment vendors need to work with industry partners to provide solutions based on differentiated requirements. While operators and vendors focus on networks and clouds, industry partners can focus on applications. 

Future of MEC: Ubiquitous Network and Computing

With the expansion of the ecosystem of network edge, more technical and service innovations will occur around edge network.
—Cloud-edge coordination: Balance between cost and performance
    The computing power of MEC has obviously higher  costs than that of a centralized cloud. The latency-insensitive services can be processed by the centralized cloud, such as massive data analysis and AI training. AI reasoning and policy execution can be delivered to MEC. Through coordination of edge and cloud, the services can achieve a balance between performance and cost.
—Edge-edge coordination: Support service continuity and trigger network edge reconfiguration
    Take V2X as an example. When a car moves, network/MEC switching will be triggered at the edge of the network/MEC coverage. To ensure the continuity of services, edge-edge coordination is required. MECs need to be connected through edge networks, thus triggering edge network reconfiguration. Edge-edge coordination includes four layers: network-network coordination, network-cloud coordination, cloud-cloud coordination, and application coordination.
—Computing-aware network: The network is aware of computing resources and capability to implement routing based on computing and network QoS
    The distributed computing power can be connected through intelligent networks so that the network can be aware of the computing power. The computing power can be allocated according to the service requirements in terms of network quality, computing power capability, and service priorities through intelligent routes to maximize the use of computing power. 
—Blockchain in MEC: Authenticate edge nodes
    Edge-edge interconnection enables the network to convert from a center-edge star structure to a multilateral interconnected structure. Blockchain technology, with features such as decentralization, consensus mechanism, and peer-to-peer interconnection, perfectly matches with the edge-edge interconnection model. The blockchain technology can be used in edge node authentication and authorization to improve service and system security.
—Grid computing: Distributed computing of huge tasks
    When scattered computing nodes are connected into a network of high computing power, they evolve into a computing grid capable of performing large/huge computing tasks. By splitting applications into several independent computing units, the network can assign them to different edge nodes and initiate distributed operations at the same time. 
—Function as a Service: Abstract general computing requirements to enable coexistence of computing and connection
    Function as a Service is an event-driven, fully-managed serverless computing service. Common computing power requirements are abstracted and then encapsulated into abundant function sets for service invocation while the service focuses only on its own logic. Function computing provides elastic, highly available, expandable, and fast response calculation capabilities for the service, so that any application or service can be quickly built.

Conclusion 

When viewing the IT development history, from mainframe computer to PC, computing power has evolved from a centralized to a distributed mode. From PC to cloud computing, computing power has evolved from a distributed to an on-demand mode. From cloud computing to edge computing, it has become more localized to enhance service quality. With the further development of the above new technologies, ubiquitous network and ubiquitous computing will eventually be realized through more intelligent network connections and more abundant computing power.