We've talked about 5G and why network slicing is an important element of 5G. Let's look at how that translates into the network architecture and what are the challenges in 5G. When we compare the 4G network with the 5G network, we see that the 4G network is fairly fixed in its architecture. There is an access network that is comprised of eNodeB's, the base stations, and then there is a centralized core network, and beyond that there is the Internet and the Cloud. This fixed architecture is very optimum to deliver the broadband services that 4G was built for. With 5G however, because we have the need to deliver different types of slices, having a fixed network architecture is not an option anymore. We need to look at specific needs of each of the slices. Whereas, the evolved or enhanced mobile broadband may have a similar network topology and network architecture as the 4G network, the massive IoT and the mission-critical IoT will have very different ideal architectures. When we look at the specific requirements that are being set for each of these slices and each of different types of services, we can see that in the case of 4G, it was fairly straightforward. There were mainly two different types of KPIs. It was the data rate per user and there was the end-to-end latency, because there was one type of service. With 5G, we see a much wider variety of requirements and use cases. We can go from the broadband access in dense areas, where downlink requirements are 300 megabits per second and uplink are 50 megabits per second, the end-to-end latency is 10 milliseconds, and mobility will range from 0 to 100 kilobytes per hour. But, there are multiple definitions of other types of use cases, where the KPIs, the user experience data rates, the end-to-end latency and the mobility will differ. This is something that really needs to be taken into consideration and we can consider each of these use case categories almost like a network slice. So, when we look at the implications of implementing different types of network slices in the network, we see that there are different aspects that need to be into consideration. On one side, there is the radio network. What is the ideal radio access technology that needs to be used for each slice? For example, for mobile broadband, the typical packet size, the typical end-to-end latency will be different than, for example, massive IoT. Therefore, the radio access technology can be different for each type of slice. In the case of the core network as well, it needs to be, some occasions you will want to have the core network closer to the user, closer to the base station, such as in low latency communications. In some other occasions, you may want to have a more centralized core network, such as the massive IoT. In some cases, you will have mobility, in some cases, you will not have mobility. So, there's softwarization of the network elements and taking advantage of software-defined networking and network function virtualization are a key enabler. Also, the separation between the control plane functions and the use of plane functions are really able to help scale in different ways for each different slice. A key aspect of network slicing is about a layered architecture, where you have a service instance layer that is responsible for the end-to-end user or business services that are supported in the network. There is a network slice layer that is responsible for the actual network characteristics required to deliver that service. Finally, there is a resource layer that is actually responsible for the physical and virtual network functions and the locations, that are required to implement a specific slice instance.
