Network Slicing means that from one network can provide different service levels. According to NGMN.org “The Network Slice Instance may be composed by zero, one or more Sub-network Instances, which may be shared by another Network Slice Instance.” To be able to provide SLA to any connection, there needs to be admission control to the slice and how to use the slice needs to be managed. This means in practice that Slice in a non-public network can’t be the best effort but capacity needs to be managed in location and time dimensions. This is well understood by industry and demonstrated in live 4G and 5G networks.
But the question not discussed yet deeply is how the Slice gets tricked, how the end user gets access to the benefits of Network Slicing. Technically there are two options: either Slice gets triggered by network or trigger is outside of network. Because of the management requirement, admission control, capacity management inside Slice and time dimension requirements means that some kind of triggering function is mandatory.
If we build a trigger inside the non-public network, it means that we’ll use Deep Packet Inspection type of technology. DPI technology is well known and contradictory, because DPI rules are not transparent. This is the reason why Net Neutrality discussion has so high tensions. This is not a problem in non-public networks but using DPI adds delay into the network. Adding delay does fit very well into the 5G race towards 0-delay connections. DPI is a common element in the existing networks, but it does not provide 100% coverage, for example HTTPS is a bit problematic and VPN kills the whole concept. This means that information and control will require more network elements and somehow all the needed information needs to be combined.
To make the system easy, it makes sense that we’ll follow Cloud principles. Let the users make the decision what kind of services they want to use. This means that the trigger to launch Slice is outside of the non-public network and the network has an API to receive requests. Just like Clouds normally have. The information that is required to launch a slice includes what kind of slice is needed and where the user is. Then the network needs to decide if the user is given the access to slice, based on network and slice loads in that location. This process is very simple and more importantly fully transparent.
To make this happen, there needs to be a customer controlled trigger outside of the non-public network and an element inside the network that can accept requests from the triggers and provide admission control to the network. This architecture does exist already in the 5G-RECORDS project for audio production.