SK Telecom’s Pursuit of MEC

The South Korean operator SK Telecom remains steady in its 5G ambitions. SK Telecom has already established a healthy foothold in the consumer market. According to the Ministry of Science and ICT, South Korea’s 5G subscriptions were at an estimated 9.25 million at the end of September 2020, with SK Telecom comprising nearly half of these subscriptions. SK Telecom plans to progress their commercial success in both the 5G consumer market and business-to-business services by placing computing functions in closer proximity to the source of data traffic through Multiaccess Edge Computing (MEC) network architecture. MEC’s distributed intelligence would mainly facilitate lower latency and higher reliability network provisioning. Processing data at the edge and bypassing the issues of transporting large amounts of data to centralized data centers can theoretically reduce applications’ and services’ final service round-trip times to as low as 10 milliseconds in a more secure manner.

SK Telecom intends to deliver its MEC services in two flavors:

• A distributed MEC architecture that aims to proliferate more cell sites that possess edge compute capabilities across SK Telecom’s nationwide 5G network. This approach will primarily target highly populated urban areas of South Korea. These edge sites will host MEC platforms developed by either SK Telecom’s in-house team or by third parties. This distributed MEC strategy will be used with standardized compute infrastructure and services.
• Its on-site MEC, on the other hand, will be focused more on private enterprises and consumers who require customized latency, security, and reliability standards. This flexibility will allow SK Telecom to cater to a diverse set of industry verticals and use cases such as smart factories, augmented reality/virtual reality, or video conferencing for online learning.

This paradigm shift in network architecture is necessitated by the unique connectivity scenarios that 5G brings—namely, enhanced mobile broadband, ultra-reliable low latency communications, and massive machine-type communications. The need for networks to accommodate lower latencies, to support higher device densities, and to effectively deliver higher bandwidth applications will require operators to optimally distribute data center computing capabilities, networking, and storage to locations closer to end users and devices. To achieve this endeavor, operators must leverage on several complementary technological innovations that can assist in their delivery of higher quality network capabilities to edge locations such as wireless access points, aggregation sites, base stations, cloud Radio Access Network (RAN) sites, fixed line access aggregation points, central offices, or even the private premises of their customers.

These complementary innovations include:

• Cloud RAN, Disaggregation of Baseband Functionalities, and xHaul: A new RAN model that supports different functional splits spread across the Radio Unit (RU) and the disaggregated baseband—comprising the Centralized Unit (CU) and the Distributed Unit (DU)—will do away with the need to collocate these RAN functionalities in the same location. This innovation will enable operators to flexibly place these RAN components in optimal locations in the network. A group of remote RUs can be aggregated into a DU via a fronthaul interface—Common Public Radio Interface (CPRI), Enhanced CPRI (eCPRI), or Radio over Ethernet (RoE)—while several DUs can be aggregated into a centralized CU through a midhaul link (F1 ethernet). The disaggregation of RAN components will dovetail with the virtualization of RAN functions on mass-produced Intel x86 servers in cloud RAN. This disaggregation (between RUs, DUs and CUs) and decomposition (RAN software running on generic, commercial, off-the-shelf servers) of RAN architecture will allow the operator to dynamically place and execute baseband workloads across different edge sites.
• Open RAN/Open Virtualized RAN (vRAN): The main objective of the open RAN architecture is how it can provide more flexibility and vendor diversity to operators in their respective network deployments by unbundling monolithic solution stacks from Tier-1 vendors. The uncoupling of hardware and software and the priority toward virtualized network functions can ideally help operators reduce both Capital Expenditures (CAPEX) and operational expenditures. Open RAN’s eCPRI, for example, will allow operators to have more options in procuring their Remote Radio Unit (RRU) or baseband unit equipment from different vendors, putting them in a better position to reduce the CAPEX required to densify MEC deployments. The traffic-flow automation embedded in Open-RAN-compliant eCPRI also allows for more efficient traffic management in the edge by reducing latency and jitter as compared with CPRI or RoE.
• Control and User Plane Separation (CUPS): As with the decomposition of the RAN into RRU, DU, and CU, CUPS architecture is based on the decomposition of mobile packet core functions. In this architecture, the User Plane Function (UPF) of evolved packet core can be situated nearer to end users to enhance network latency and reliability while the Control Plane Function (CPF) can remain centralized. This architecture is the foundation of MEC as CUPS allows for the geographical distribution of multiple UPFs to the network edge while the CPF can be located in centralized data centers. CUPS will enable MEC to support both network infrastructure functions (i.e., SK Telecom’s Distributed MEC) and a variety of low latency, stringent quality-of-service applications—from content delivery to supporting mission-critical use cases (i.e., SK Telecom’s On-Site MEC).

Operators making the transition to MEC services should first start by implementing robust MEC infrastructure at the network edge before progressing to more specific customized applications and services. To successfully achieve impactful end-user/enterprise use cases, such as smart factories, mission-critical healthcare, or financial services applications, operators will need to first focus on foundational MEC network elements such as such as Cloud RAN, vRAN, CUPS, Remote PHY, and so on. It is through these infrastructural building blocks that future applications can come to fruition.

Further, operators must be aware that there are no standardized edge architectures. Multi-stakeholder edge environments offer an expeditious route to achieving the objectives of MEC depending on the resources or desired use cases. In these instances, hyperscalers might provide the underlying core infrastructure while hardware vendors (i.e., Intel) and software providers (i.e., VMWare) contribute their respective components. Operators can then tap into these edge cloud resources by integrating them within their MEC networks. Take for example the recent collaboration between SK Telecom and Amazon Web Services (AWS) and their launch of “SKT 5GX Edge.” SK Telecom is leveraging on the embedded AWS compute and storage infrastructure within the edge data centers of its network (referred to as “wavelength zones”) to deliver a platform that enables customers to build mobile applications requiring ultra-low latency.

SK Telecom’s partnership with AWS demonstrates the imperative of having an adequate supply of foundational MEC infrastructure to achieve more sophisticated use cases. The partnership is a testament to how operators and hyperscalers can form a symbiotic relationship in large-scale MEC deployments. Hyperscalers have more familiarity with application and device ecosystem development, more resources and expertise in providing compute and storage capabilities from their existing data centers, and more experience in providing cloud services to enterprises. On the other hand, operators have existing infrastructure assets such as cell tower sites, central offices, and transport networks. Operators would also have more expertise in delivering 5G network services such as network slicing.