5G | Broadband/Network | Emerging Trends
July 4, 2017

Why end-to-end network slicing will be important for 5G

By Peter Ashwood-Smith

If you have been following developments in telecommunications in the last year or so you have no doubt heard the term “slicing” as it pertains to future 5G networks. This short article will hopefully give you a high-level view of what “slicing” is, why it’s important for 5G systems, and some indication of the work accomplished by the ITU Telecommunication Standardization Sector (ITU–T) on this important new technology.

The need for slicing future network systems can perhaps best be understood by looking at city transportation systems. In a city, we don’t provide a single transportation mechanism. Instead, the infrastructure of the city is divided — sliced, if you will — into areas for cars, buses, subways, etc.

Some of the infrastructure is dedicated to a par­ticular form of transportation (eg. trains), while other infrastructure can share different kinds of transportation (eg. roads are shared by cars as well as buses, which may have priority lanes).

This analogy mirrors nicely what we plan to do with 5G. Essentially, we intend to take the infra­structure resources from the spectrum, antennas and all of the backend network and equipment and use it to create multiple sub-networks with different properties.

Each sub-network slices the resources from the physical network, end to end, to create its own independent, no-compromise network for its preferred applications.

“Much of the challenge with 5G will be providing the proper degree of orchestration that ensures harmonious end-to-end operation.” – Peter Ashwood-Smith, 5G Network Research Director, Huawei

Next-generation network challenges

In today’s Internet-of-Things (IoT) era, we are creating new types of machines, both big and small, at an amazing rate. Connecting these machines offers great opportunity, but brings with it a host of challenges.

Today’s 3G/4G/LTE networks do a wonderful job connecting people, but they pose a number of problems when used to connect machines. This is because the 3G/4G/LTE networks were designed as a set of compromises.

For example, the 4G/LTE networks do not give the lowest possible delay because to do so would have an adverse effect on the band­width they could provide. Likewise, the careful scheduling of individual users through multiple message exchanges creates higher throughput and more fair access, however, it uses consid­erable battery resources in the handsets to do so. Some of the challenges for next-generation applications and the current situation with 4G are depicted in the figure.

Slice types

To address the different needs of different types of machines and devices, the interface between the device and the antennas (the air interface) will have several different specialized/ tailored behaviours. These are referred to as slice types.

One slice type is specifically targeted for ultra-low latency and high reliability (like self-driving vehicles) (URRLC), another slice type is spe­cifically targeted for devices that don’t have large batteries (like sensors) (MMTC) and need efficiency and yet another slice type is targeted at ultra-high speed (eMBB) as required for 4K or immersive 3d video. While the initial standards work calls for only three slice types, the architec­tures are flexible for future slice types.

Since it would be far too expensive to allocate a complete end-to-end network to each type of slice, the network infrastructure that supports 5G (and likely 4G) will employ sharing techniques (virtualization and cloud), which allow for mul­tiple slice types to co-exist without having too many multiples of the resources.

Cloud and packet-based statistical multiplexing techniques are employed to allow the slices to use each other’s resources when they are free. In this manner N-network slices can be imple­mented with far less than N x the number of resources. This is depicted in the figure.

In order to make such networks a reality all the components need to work properly in harmony. Much of the challenge with 5G will be providing the proper degree of orchestration that ensures harmonious end-to-end operation, and as a result, this is one of many areas under study by ITU–T.

ITU–T − researching standards for 5G

ITU–T’s Study Group 13 (SG13) recently cre­ated a Focus Group with a mandate to research the areas that needed standardization for the non-radio aspects of 5G. The harmonious oper­ation through software control, referred to as “softwarization” of all of the components of the 5G network, was one of the many subjects stud­ied by the Focus Group, and which is now being more formally considered by SG13. Many of the areas requiring control are not uniquely wireless components but are also involved in service providers’ other end-to-end-businesses.

For example, the cloud and transport networks which interconnect them will require new agile control to ensure that the packet, non-packet interconnections and compute, meet the Quality-of-Service (QoS) demands of that slice.

The success of 5G lies in entire ecosystems

5G slicing technology, to be truly successful, will need entire ecosystems to come together to solve and standardize their end-to-end applications.

As a result we fully expect to see the automo­tive, health care, agricultural, manufacturing etc. ecosystems to become more and more involved in 5G and to help drive the potential that slicing can provide.

Peter Ashwood-Smith
ITU–T IMT-2020 Focus Group Chair and 5G Network Research Director, Huawei
[Read the latest edition of ITU News Magazine to learn more about the transformative potential of 5G — and how ITU is helping to make it happen.]
  • Was this article Helpful ?
  • yes   no
© International Telecommunication Union 1865-2018 All Rights Reserved.
ITU is the United Nations' specialized agency for information and communication technology. Any opinions expressed and statistics presented by third parties do not necessarily reflect the views of ITU.

Why end-to-end network slicing will be important for 5G

Send this to a friend