5G: How Many Slices are Enough?
The success of 5G depends on the ability to support a diverse set of services with varied requirements for latency, bandwidth, density and availability.
The 5G network needs to be highly adaptable to support new services and business models that will emerge with 5G. The best effort, “one-size-fits-all” approach to mobile networks used in previous mobile generations will no longer be viable.
So network slicing was proposed to allow a single network to support multiple different service types. Basically the network is cut into multiple slices and each slice is assigned a set of the network assets. So imagine cutting a pizza into slices and giving each slice to a different person, some might have olives with their slice, others might have mushrooms and still others eat it plain. It is key that each slice is isolated so that one slice cannot take resources allocated to a different slice.
Taking the pizza example again, once you have consumed your slice you cannot take a slice from another person without causing issues for someone else. It should be noted that an individual asset can be shared across multiple slices as long as isolation is maintained. So for example a single compute asset offered by the network maybe used by multiple slices, with one slice running two virtual machines (VMs) and another running three VMs as long as the amount of compute resource per VM is protected. Or one slice may require 100G of node capacity and another may require 200G, this is fine as long the capacity is reserved. Once a slice is created it behaves like its own independent network and services can be provisioned across it in the normal fashion.
So how many slices are enough?
Too few slices and the range of services supported on each slice will just be too diverse to make the slice viable without over allocating capacity, connectivity and compute across the network to support the service mix.
Too many slices and the network is just cut too thin to be of operational use. Think again of the pizza, there comes a point where you can’t cut it any thinner, and even if you could can it hold the sauce you want?
So for the sake of simplicity let’s assume that there will at least be a slice per broad service type identified above/below:
- Superfast Mobile Broadband
- Basic Voice Services
- Ultra-reliable services
- Ultra-low latency services
- Mission Critical IoT
- Massive IoT Services
Then in some cases, it makes senses to create a specific slice for a specific user. So for example, a large enterprise like ‘Acne Burger’ or a strategic industry like ‘Spring A Leak Water plc’ might require a set of communications services. For security reasons, they might demand a private network for their services. ‘Acne Burger’ might not want their services shared on the same slices as “Cheaper Burger” and “Spring A Leak Water plc” might want air gap security between them and ‘We Bring Down your Infrastructure.com’. However, if slice isolation can be proven, and the slices can created on non-shared assets, then they might be confident that they have a private network. In addition, it is then relatively straightforward to provide each enterprise with a portal that allows them to monitor their services and the network slice they are transported on.
For the first time, network slicing gives MNOs a real opportunity to provide service differentiation between them and their MNO competitors. The MNO can decide to target certain market segments and put network slices in place with assets optimized to support these services. I already mentioned the ability to give enterprises and strategic industries a portal to manage the lifecycle of their own services and slices. The MNO might target specific sectors, for example if it targeted virtual reality mobile gaming it could put a slice in place specifically for this, rather than a more general low latency slice. This slice could then be created with assets that allow it to be optimized for response times, reliability bi-directional capacity etc. In the extreme, the MNO could tie-up with the gaming firm and produce a slice with deterministic characteristics that the game designers built around when developing the game !
So how many slices are enough. The right number to support your business model.