Edge computing has been attracting a lot of interest lately. The reasons are many, not the least is the potential for service providers to create and monetize new services and applications. I recently chaired a panel on MEC at the Layer123 NFV Congress 2017 in San Jose. In speaking with different ecosystem players, I noted a few misconceptions related to edge computing:
- The edge is at the base station: deploying edge computing at the base station is one of the most common misconception. It is also the most expensive of options. Moreover, it does not yield the best performance in many applications. In the context of cellular networks, service providers will not opt for this option. Instead, they seek aggregation points elsewhere. Some of the new architectures include distributing the core network, bringing it closer to the user which brings a whole different and new dimension to edge computing. On the other extreme, the end device is the edge and aggregation at the end device level makes sense in IoT applications. This is one of the premises of ‘fog computing.’
- Virtualization is a pre-requisite for edge computing: Virtualization of the mobile network would accelerate the implementation of edge computing, but edge computing is not strictly dependent on virtualization of the radio access network (i.e. C-RAN, vRAN, etc.), or the core network (e.g. vEPC). Virtualization provides a flexible environment to implement edge computing servers and applications and it is taken for granted that such an environment is virtualized.
- ETSI MEC is the standard approach to edge computing in mobile networks: ETSI is in process of defining and specifying the architecture and interfaces for edge computing in mobile networks, but this is not a standard that vendors must comply with. ETSI MEC does not define all interfaces: some important ones are left to proprietary implementation. Nokia’s Liquid App solutions are the precursor to ETSI MEC where Nokia still plays an active role. Vodafone had highest interest among MNOs and other vendors, such as Huawei, joined the initiative as interest heightened. Ericsson on the other hand, does not subscribe to the same philosophy and does not participate ETSI MEC. Recent developments at ETSI MEC show a broadening of scope in many aspects such as architecture (e.g. Gi interface in addition to S1 interface implementation), and technologies (e.g. incorporate Wi-Fi and fixed access technologies). This is in line the evolution of service provider requirements. There are other initiatives that incorporate edge computing inherently such as M-CORD.
- Edge computing is for 5G networks: Edge computing is just as applicable in 3G or 4G networks as it is in 5G networks. In fact, 3G and 4G would arguably benefit from edge computing more than 5G networks which offer lower latency performance. The election to implement edge computing is primarily a business decision taken to enable new services and generate new revenues.
- Edge computing is the domain of mobile service providers: This is an issue of business models and go-to-market strategy. Edge computing allows new service providers who are not mobile network operators to enter the market. Granted that collaboration with the MNOs is mandatory. However, rising opportunities will enable a new class of service providers – neutral hosts – to operate their own local networks. This is achieved with small cells in shared spectrum (e.g. 3.5 GHz CBRS band) or in unlicensed spectrum using MuLTEfire technology. Edge computing will enhance the value proposition of such a class of service providers. In fact, it will be a long ramp to implementing edge computing if the concept remains the sole preserve of the MNOs. This is because edge computing enables services in the domain of CIO offices and IT departments. Telecom service providers are generally weak at serving the enterprise: this is not ‘selling SIM cards.’