What is network slicing?
Courtesy http://keithsmarketingsite.com/ Courtesy: https://jameswoo.net/diy-dry-box-how-to-make-a-dry-box
Take a look at the two camera cases above.
Figure 1 has compartments for camera and each of the accessories. Every accessory is different in height, width, length and the structure as a whole. This case makes a distinction between the accessories and has a separate compartment custom made for each type.
The camera case in figure 2 uses a One-Size-Fits-All architecture. It is big enough to fit in the camera and all the accessories, however it doesn’t distinguish between any of them. The case doesn’t really care what goes in, and as long as there is space any number of accessories can be accommodated.
Mobile communications, before the advent of 5G, worked like the camera case of figure 2. The Radio or the Core network didn’t make any distinction between the devices that connect to the network. The same RAN and Core network will service all devices.
But in reality, many kinds of devices connect to the network with varied properties.
Type of devices:
Smartphones, tablets and phablets constitute a major portion of the mobile devices today. As per Cisco VNI studies, it is expected that by 2021 there would be around 8.3 Billion handheld or personal mobile devices. These devices are not only mobile, but can also consume a high volume of data. Network latency is generally not a prime concern here. With enhanced Mobile Broadband (eMBB), 5G can provide high speed internet in a small area. Devices in that area will generally be static or have very low mobility but would require a high data bandwidth. For e.g. 5G may be used to provide high bandwidth to a company office which is currently using fixed line broadband.
The Internet Of Things (IoT) devices have been increasing steadily over the past few years and it’s expected that by 2021 there would be as many as 14 billion of such devices. These devices are low powered and send very little data to the network. They are fixed to a location (like application sensors in a industry ) and have very low mobility. However, number of devices that would attach to a cell would be far greater than any mobile device. Consider your work place – the lights, air conditioners, doors, windows, projector in the conference room and many other things could potentially connect to internet as part of connected workplace. Even in a small area, hundreds and thousands of devices may connect to the network
The need to improve road safety and development of driverless cars has encouraged the growth of Vehicle to Everything(V2X) communication. The vehicles on the road are expected to communicate to other vehicles, pedestrians and infrastructure. V2X( more specifically C-V2X or Cellular-V2X) is being extensively tested by mobile operators and automakers and initial deployments are expected towards the end of 2018. The data traversing through the network will be less, but by their very nature, vehicles will be highly mobile. The network will have to handle high mobility with low data bandwidth.
Some devices require a very reliable network with extremely low latency. Devices which are used in telesurgery, autonomous vehicles, augmented/virtual reality etc. fall into this category. These devices are generally not very mobile but are expected to consume massive data. These devices require tight delay and extremely low end-to-end round trip times of < 1ms. They require extremely reliable network connectivity, as even the slightest error could lead to catastrophic outcomes.
In 4G and all mobile communication technologies before that, all mobile devices are treated equal. With a few features such as preemption capabilities and QoS there is an attempt to cater to all the various types of devices (One Size Fits All).
5G understands the unique needs of each device and provides a way to create a separate network for each type. This concept is known as network slicing.
Network slicing allows the operator to provide dedicated logical/virtual networks for specific requirement and functionality each having their own unique properties. The services are abstracted from the physical resources.
How does network slicing work?
Network Slicing uses the concept of Software Defined Networking (SDN) and Network Function Virtualization (NFV). Each network slice may have its own network architecture, protocols and security settings. Network slicing includes slicing of Radio Access Network (RAN), Core Network (CN) and maybe even the end devices. The figure below depicts how various network slices can be used to provide distinct QoS to each type of device.
Each slice has a separate RAN and Core slice. Let’s take a look at the unique features provided by the above slices.
The eMBB slice could be the default slice allocated to the subscriber on attaching to the network. All internet/data traffic from the subscriber passes through the RAN to the Core and finally reaches the application server.
The URLLC slice steers the traffic to the edge compute. The edge compute is located very close to the subscriber thereby reducing the network latency to less than 1ms.
The IoT slice has a different kind of radio. IoTs require high subscriber capacity but very low throughput. Thus that slice is tailored for IoT devices only.
3GPP Control Plane User Plane Separation (CUPS) architecture can be used to efficiently fulfill the distinct requirements of the various network slices. To reduce the network latency for the URLLC slice, the user plane function can be placed very close to subscriber – in the edge cloud. Similarly, an user plane function that can handle high subscriber capacity can be placed in the IoT slice.
End-to-End network slicing will include the Core, the RAN and possibly the transport network as well.
RAN Network Slicing: Ran slicing is important to achieve sharing of radio resources. This is enabled by a Software Defined RAN or Cloud RAN. The controller in the cloud RAN can allocate appropriate computational, network and radio resources to various network slices based on the their service requirements.
Core Network Slicing: It’s relatively easier to implement SDN and NFV on core network . Based on the service type necessary VNF can be provided for the network slice. These VNFs can be scaled on demand with changes in service and performance requirements.
Transport Network Slicing: Considerable effort is being put to improve the backhaul to satisfy the 5G needs. Leading vendors have already demonstrated network routers that support backhaul slices with independant performance characteristics. SDN will play a crucial role in determining the best path for a packet to traverse through the fronthaul and backhaul elements, to meet the various 5G demands.
Benefits of Network Slicing
Optimized Resource Allocation: Network slicing enables creation of a virtual network fine tuned for the service. Each slice will have only those network elements that are relevant for the slice. For e.g. Let’s say the operator decides to charge IoT devices based on the number of devices only and doesn’t care for their data usage (which is expected to be very small anyway). The IoT network slice, in that case, need not have an OCS network function.
Cost Reduction: For network slices that do not require heavy computation (like the IoT network slice), operators have the luxury to reuse the legacy older hardware having low compute capacity. This could potentially drive down the operating expenses significantly.
Network Slicing Use Case
Let’s consider a simple use case to understand network slicing.
Bob buys a new sim and attaches to the operator network On attaching he is assigned the default slice. Default slice transmits data on a best effort basis. Very soon the network gets congested and Bob doesn’t get the same network speed as he used to get earlier. He upgrades himself to a premium subscriber and gets a dedicated network slice. Dedicated network slice routes the traffic through the edge cloud which reduces network latency and gives Bob a good video experience.
A picture is worth a 1000 words ! 🙂 . Take a look at the story below.