Among its other capabilities, the 5G network supports slicing, a technique that divides a single physical network infrastructure into multiple virtual networks. Compared to existing 4G and LTE (Long Term Evolution) networks, 5G promises significant improvements in bandwidth and latency, making virtual network slices a possibility.
Each virtual network instance created by 5G network slicing provides an isolated, end-to-end network, optimized for a specific business purpose. These virtual networks can support a wide range of services and applications, which fall into three general categories:
- Enhanced mobile broadband. This high-bandwidth cellular service includes voice and SMS. Think of this as more bandwidth for your cellphone to enable applications like 4K video resolution and augmented virtual reality, which need throughputs of 10 Gbps and higher.
- Ultra-reliable and low-latency communications. Autonomous vehicle inter-communications is the typical use, where fast, highly reliable communications between self-driving cars is required.
- Massive machine-type communications. This includes IoT applications for wireless sensing and control devices, which might be found in a factory.
5G network slicing architecture is described further by the 5G Infrastructure Public-Private Partnership Architecture Working Group. The architecture uses a recursive, multi-tenant model in which an infrastructure provider's physical network is divided into sub-networks.
Each sub-network is leased to mobile virtual network operators, which divide the allocated network into more specific sub-networks. Each sub-network in turn is bundled with cloud services encompassing compute and storage infrastructure to meet a specific circumstance (see image, "How 5G networking slicing works").
Business purposes and applications dictate the mix of bandwidth, latency, resiliency, processing and storage required by each slice. Content distribution, IoT edge computing and network functions virtualization influence the mix of compute and storage within each slice.
Slice control and management
5G network slicing management must include the ability to create, operate and delete slices. These steps must be automated for quick and accurate control. The same system abstractions are used in each layer, regardless of whether the underlay is based on physical infrastructure or a logical network slice, facilitating the use of automation for managing slices. In addition, automation allows IT to rapidly provision new slices and remove them when they are no longer needed.
Control and management systems will use common APIs to enable oversight of network slices, but physical infrastructure providers will still need dedicated tools to control and manage parts of the underlying physical infrastructure. While a virtualized network slice can detect and report network errors, it won't be able to identify the physical infrastructure component that's the culprit.
Network security is another consideration. There are trust issues between the delegated operator of a sub-network slice and the owner/operator of the parent slice. Traffic in a slice also has to be segregated from other slices, much like we have today with virtual routing and forwarding instances.
5G aggregation, or the opposite of slicing
Another offshoot of virtual slicing is 5G network aggregation. Consider a situation where a mobile network service supplier has to provide service that spans beyond a single physical infrastructure provider's network. Perhaps one provider offers exceptional coverage in part of a city, while another offers equal coverage in other areas of the same city. An IoT sensor network slice used to monitor vehicle traffic flow could be built by aggregating slices from the two providers. This design would allow the IoT sensors to communicate directly with edge computing systems instead of being required to transport a larger volume of raw data all the way back to the application servers.
Network slicing brings a new twist to networking, and several vendors have showcased how the technology will work. The next step is to determine how it performs in the real world.