Edge computing is the key enabler for 5G and its cloud-based structure. The experts of CNIT (Italian Inter-University Center for Telecoms) explained why at the 5G Italy event.
This new approach is very challenging to explain, so let’s start with a short list of bullet points. Andrea Detti (University of Rome – Tor Vergata) summarized the technical similarities between cloud and 5G in three key points:
- 5G, which splits radio and processing units, has a lot in common with the structure of the Web itself;
- 5G no longer relies on nodes, but on functions: it needs underlying cloud computing;
- more technically, 5G-related software uses a Rest-based approach on open APIs that are ready to use on Github through Docker containers.
A technical paper was presented at the 2019 5G Italy conference by Fabrizio Granelli (University of Trento) and Riccardo Bassoli (University of Dresden). It helped to explain the importance of 5G architecture through the radio-access network and edge computing.
5G will incorporate several state-of-the-art architectural and protocol approaches on top of its networking infrastructure in order to tackle the emerging need for improved flexibility and performance.
Such requirements lead to the emergence of two design options for its architecture: an evolution of 4G-LTE standard IP architecture, and interaction of the core network functions using a Service-Based Architecture. This novel architecture allows for the integration of the recent developments in cloud and mobile edge technologies, as well as leveraging energy efficiency and latency.
5G application scenarios include vehicles and robots
The key approach to the coming revolution is Network Function Virtualization. 5G is expected to support unprecedented requirements including very low latency (in the order of 1 ms), high reliability, and the capability to offer access to distributed computation or storage facilities. These new characteristics are in addition to higher connectivity and bandwidth.
Several target scenarios and services include strict constraints in terms of low latency and extreme reliability, classified as URLLC (Ultra-Reliable Low Latency Communications) services. Vehicular communications and remote control of robots or machinery belong to this class of services and are well-known 5G application scenarios.
By virtualizing the key network, it is possible to detach software functions from dedicated hardware, with the advantage that one is then able to relocate or modify resources very efficiently and in real time. This emerging paradigm is what we define as NFV, Network Function Virtualization.
NFV leads to the definition of two next-generation paradigms: C-RAN (Cloud Radio Access Network) and MEC (Mobile Edge Computing/Cloud). Let’s take a closer look.
The Cloud Radio Access Network
Cells are becoming ever smaller in order to provide high performances. This creates problems in terms of both the cost of the equipment itself and its power supply. C-RAN represents a possible architecture with which future cellular networks will tackle cost and energy consumption.
Base stations are designed to handle the maximum traffic possible, not simply average traffic. This results in a waste of processing resources and power at idle times or when there is a low level of traffic.
The majority of power sources don’t scale according to the number of users or the amount of traffic. Therefore, a more flexible solution is required.
The typical 4G RAN architecture in the eNodeB architecture contains both RF and processing capabilities. Cloud RAN separates the two components, putting processing units in a BBU (baseband unit) Pool and connecting to the RRH (Remote Radio heads). BBU pools are based on open platforms, and real-time virtualization technologies rooted in cloud computing achieve dynamic, shared resource allocation and support for a multi-vendor, multi-technology environment.
Cloud RAN enables large-scale centralized deployments, allowing hundreds of thousands of remote RRHs to connect to a single centralized BBU pool.
A detailed end-to-end mathematical framework has recently been developed at the CINT research unit in Trento, in order to study the potential benefits of 5G.
Mobile Edge Computing, or MEC
MEC is a cloud-based IT service environment that runs applications closer to the customers. This way, latency and network congestion can be reduced, giving generally higher performances.
A MEC application is hosted in the distributed cloud of a MEC system. The application can belong to one or more network slices (i.e. virtual networks) that have been configured inside the 5G core network.
Cloud RAN and MEC represent a big step forward in the flexibility of the 5G network infrastructure, increasing scalability and service support.
Edge computing will be central to 5G, added Maurizio Dècina, president of Infratel, during the main conference. “If there are about ten core centers, we already see hundreds of regional data centers and thousands of other dedicated data centers, a very large number for a nation like Italy”.
Edge computing is nonetheless the key, in terms of both data ownership and performance. “The edge computing approach, in fact, makes it possible to exploit the network at 70/80%, where wasting huge amounts in core data centers would be enormous”.