Network slicing: a next generation 5G perspective - EURASIP Journal on Wireless Communications and Networking

The 5G-NSUE model has been developed after undertaking a comprehensive review of past and present network slicing systems. The proposed system has been built by utilising software and virtualisation techniques on the underlying network infrastructure to create network slices according to different use cases. With the help of domain experts in SDN, 5G networks, virtualisation and network slicing, the classification was refined according to the most relevant factors for the design, modelling, validation and evaluation of the network slicing system. On the basis of the review and prior knowledge of the field in order to create slices in the 5G network, four main points need to be considered:

1.

selection of network area on which network slicing should be implemented on;
2.

the scope of the slices of the 5G network;
3.

how the network slices should be delivered to different use cases, and
4.

which technologies should be integrated with the network slicing model.

These factors are crucial elements of network slicing architecture and will help provide an understanding of the requirements that should be considered for creating any network slicing architecture on 5G networks. Moreover, these factors are classified with their attributes to provide an in-depth understanding. Previous works studied several of these factors in their architectural model, but did not consider all these factors at the same time. Moreover, previous state-of-the-art models considered using either only SDN or NFV in their architecture. By contrast, the proposed model considered these factors, which are equally important for network slicing. Given that the previous models considered typical cases, such as core network slicing or RAN slicing, the proposed model considered slicing on the core network, RAN and user equipment, which constitute the complete network slicing of the 5G network.

The first considered point is the area for the slicing model in the 5G network. This network area can be divided into three main components that are RAN, core network and the user equipment. These components are the important aspects of the network node that should be considered when deciding which area of the 5G network where network slicing will be implemented on. The main 5G network components include underlying physical infrastructures, such as base stations, end devices, switching centres and mobility management units. These components are important because of the possibility of creating slices of the 5G network in these individual components and in the combination of them. The second considered factor for slicing for slicing model is the network slice scope or the network slicing step, in which the overall slicing processes are performed. The subclasses in this factor are slice creation, isolation and management. As evident by the name, aforementioned components are useful for creating and managing the network slices in a 5G network. This step can be termed as the key part of the system model given that this factor involves the main task of the research. Moreover, the SDN controller is further divided into infrastructure and tenant SDN controllers.

The third factor is 5G network node and considered use cases of the network. The network nodes are subclassed into three groups: RAN, core network and user equipment. The bases for the 5G network are the different use cases that the network will serve. To serve different use cases, the 5G network will use network slicing to create slices with different attributes that are required by the multiple use cases. The importance of this factor is to understand the requirements and be able to create the network slices to feed their necessity and demands. The subclasses in this factor are mobility, resource management, security, low latency and high and low bandwidths. These subclasses show the nature of requirements that need to be facilitated by the network slices. The fourth and final factor in this system model is the use of virtualisation and software techniques. These integrated techniques form the foundation for network slicing in the 5G network architecture and are integral factors that cannot be missed from the system model. The subfactors are SDN and NFV. SDN is simply an abstraction for describing components and functions, as well as the protocols for managing the forwarding plane. This system model concept using NFV emphasises the use of virtualisation for various network node functions.

Table 1 shows the components and sub-components of the proposed network slicing model. The columns of the table present the main attributes of each component and their sub-components along with several common instances for each case. The components column includes the name of the components required in the model while the main attributes column includes value/feature/function for the respective components. Instances are the generic examples for the respective components. Following the table is the component diagram, Fig. 3 illustrates the relationship between these components and their sub-components and how they are linked with each other. This figure shows different components of the 5G-NSUE system model. The four units indicate four different components and their sub-components in the system. The component and sub-component are connected using dotted lines. The solid lines are used to communicate between multiple components or to and from sub-components of the same factor/component. Integrated techniques comprised sub-components used to softwarise and virtualise the physical infrastructure of the 5G network. This step is achieved by using SDN and NFV, which are applied on physical infrastructures such as RAN, core network and user equipment. User equipment is connected to the network via RAN, which is connected with the core network. The core network helps connect with third-party networks, such as the public Internet. By using the virtualised and softwarised components of the network, the network slicing component will cut the slices of the network. The slices are created according to the demands and requirements from the different use cases that are connected with the network. These requirements are received when the request for slices is made by the use case. After the slice is created, the isolation of these slices must be considered depending on the type of slices. This step is performed by the slice isolation sub-component, while slice management component is responsible for the overall management of the slice orchestration and management of network slices. The remainder of this section will define each of the four factors and their subclasses and justify why each factor is used for classification. Diagrams of the classes and subclasses, which make up each of the factors, are illustrated accordingly. Table 1 shows a component table for the network slicing model and shows the required components along with its sub-components. Each component and sub-component are further provided with their attributes and instances in other columns of the table.

Table 1 5G-NSUE proposed system components

Full size table

Fig. 3

Proposed network slicing model. Dotted lines show sub-components whereas solid lines depict an action or process coloured solid line are used to show which process is involved with each component/sub-component. The process belongs to the sub-component from which the solid line starts from. For example, blue solid line belongs to the SDN sub-component as the blue solid line starts from SDN sub-component and ends at RAN sub-component. Blocks are used to group each component and its sub-component to a single unit

Full size image

4.1

Network node

Network slicing can be created on different areas of 5G network. An important factor is to consider the network area where network slicing is to be implemented. Before the advent of 5G technology, the core network and RAN did not distinguish between the devices that connected with them. The same core network and RAN served all the devices, hence the need for clear distinction between these areas of the network. Network slicing can be implemented on different areas of the network, such as the core network, RAN and user equipment. The different reviewed studies implemented network slicing on these different areas of the 5G network. Several articles have been found that have implemented network slicing on the core network and the RAN, but few instigated network slicing on the end user equipment. Most studies applied network slicing on individual areas only and not combined, such as the core network or RAN only. Carefully selecting the area of the network on which the slicing is to be implemented is important. The most effective configuration for network slicing is the combination of all the three areas, where network slicing is performed at different network nodes and areas.

Figure 4 shows a design architecture of 5G network with different nodes. The main network nodes of the 5G network are RAN, core network and user equipment. The RAN connects the user equipment with the core network. The RAN comprises connected base stations with controllers. For 5G, the RAN (also called NG-RAN) base station (also called gNB) has three main function units: The Centralized Unit (CU), the Distributed Unit (DU), and the Radio Unit (RU), which can be implemented in various combinations. The core network is also incorporated different units, with each unit responsible for performing different functions.

Fig. 4

Network nodes of the 5G network architecture

Full size image

The control and user planes are separated to distinguish between the user plane function and control plane function. Different control plane functions include authentication, policy control, access and mobility and session management. The core network can be connected to third-party networks, such as cloud servers, via the Internet. The purpose of using a network node as part of the model will identify the part of the 5G network where network slicing will be executed/implemented. In addition, this step will determine the extent of the network slicing. For example, network slicing can be performed on only one area of the network or on a combination of multiple areas of the network. As the nodes of the network in which network slicing is increased or combined, the complexity of network slicing will increase. However, doing such might create network slices in multiple nodes and thus complete network slicing on the 5G network, which will increase the performance of the slicing. The subfactors involved in this classification are user equipment, RAN and core network. When we think of network slicing, we mainly focus on core network slicing. To extend this definition, we can also think of network slicing in RAN node. However, performing network slicing on the RAN seems more complex and difficult than performing network slicing on the core network only. Moreover, network slicing can be implemented on end devices or the user equipment. However, the problem with network slicing in user equipment is in its initial phase and network slicing in user equipment seems to be a far-fetched concept at this moment. Nonetheless, the location is still a potential area, and achieving network slicing on user equipment will enhance the overall performance of the 5G network.

4.2

Network slicing

Network slicing can be defined as the technology that enables the creation of different virtual networks on top of physical infrastructures. Network slicing is the central figure in the system model. This component will help create network slices according to various demands, which come from another component of the system model, namely the use cases. Different use cases exist. Thus, the requirements from these use cases are different in nature. These varying demands needs to be fulfilled by today’s 5G network, which is provided with the proposed network slicing component. The Integrated technique component will create a virtualised network scenario in which the physical infrastructure of the 5G network is secondary, and the logical components that are the abstractions of the underlying physical infrastructure are of prime importance and are primary components. These primary or logical components can fulfil specific purposes according to the need from the use cases. The important aspect to consider in this situation is that the logical network is adaptable and can make adjustments according to the changes in needs by devoting more/less resources in the process. With network slicing, the 5G network can now be deployed more quickly given that only fewer functions need be deployed according to the use case (unlike when all functions are being deployed) and the users utilise only what is needed. This network slicing component is dependent on the network nodes, which include the core network, RAN and the end device. The reason for this dependence is that network slicing needs to be implemented on these infrastructures. Network slicing can be deployed on the core network only, on the RAN only or on the combination of both. The sub-components of network slicing involve slice creation, which enables the development of 5G network slices; slice isolation, which isolates the different type of slices from one other; and slice management, which, as the name suggests, manages the overall process from slice creation to slice delivery to the use case.

Figure 5 shows three different layers: the resource, network slicing and service layers. The resource layer consists of the physical resources or the network functions, which are used to provide services to the end users. Before services are delivered to the end users, the resources are first sliced to create different instances called subnetwork instances, which are then utilised to form the network slice instances. Several subnetwork instances may coalesce to form a single network slice instance as depicted by the colour of the blocks. Alternatively, a network slice instance may be directly delivered from the network functions and shown with Network Slice Instance 1, which is created directly from the network function rather than by using subnetwork instances. Finally, the network slice instance is used to create the service instance, which provides specific services to the end users. These service instances are created per the demands of different use cases. The purpose of using network slicing as part of the model is to identify the aspects of network slicing used. For example, most of the reviewed literature uses the slice creation aspect of network slicing, but not all of them considered the slice isolation aspect of network slicing. Accordingly, we used network slicing as part of the model. Moreover, network slicing is the core of this paper’s research topic. Thus, classifying prior studies according to network slicing seems obvious. Different aspects were used as part of the model under network slicing, including slice creation, slice isolation and slice management. Slice creation is mostly used in all the extant research, while few selected literature focused on slice isolation and slice management. Considering the aspect of slice isolation and slice management is crucial because these aspects will identify the efficiency and effectiveness of network slicing; thus, they are used in the model. Creating isolation between the different types of slices is also vital so that they can be delivered to different use cases as required. Furthermore, slice management will ensure that the overall process of network slicing is performed in a controlled way and the slices are processed through slice management and the orchestration unit. The problem with slice management and slice isolation is that they require advanced computation and processing which will increase the model complexity; however, the inclusion of these aspects will possibly ensure the complete package of network slicing.

Fig. 5

Network slicing model in a layered approach

Full size image

4.3

Use cases

Many use cases must be served by the 5G network. The 5G network needs to be sliced according to the varying requirements of different use cases. The emergence of IoT has added to the brackets of use cases that the 5G network must serve. Network slicing must be implemented such that these use cases are delivered with the right network slice according to their demands and requirements. Furthermore, the dynamic nature of use cases adds to the complexity of network slicing. Network slicing must be flexible and dynamic enough to sustain the needs of the changing nature of the requirements of use cases. Use cases have different attributes, such as ultra-high bandwidth use cases, very low-latency use cases, ultra-reliable low-latency use cases, high-bandwidth use cases and massive IoT use cases. Massive IoT [25] is one of the most anticipated use cases of the 5G network and will require the sliced form of the 5G network. It uses the sliced 5G network to seamlessly connect different embedded sensors all over the world. The attributes of such use case include smart cities, assets tracking and smart utilities up to agriculture industries. Ultra-reliable low-latency use cases use highly available low-latency links for purposes such as remote control of critical infrastructures, smart grid control, the automation of industries, robotics and drones. Enhanced event experience use cases have attributes that are related to VR videos, as well as high-definition and high-fidelity media experiences.

Figure 6 shows a simple demonstration of different use cases to be served by the 5G network slices. Different colours show different types of slices, and slices with similar requirements are grouped together. Different use cases require different types of network slices and thus, the created slices serve multiple use cases, as shown in Fig. 6. The purpose of utilising use cases as part of the classification is that the use cases will help differentiate the type of slices that will be created. Numerous use cases must be served by the 5G network, and all these use cases require different types of network services and demand different natures of network slices. Thus, with the use of the slicing algorithm, different types of slices can be created and delivered to these use cases. Use cases will help identify the nature of slices that must be created by the 5G network. Thus, considering use cases as part of classification was vital. Different subfactors under use cases were used for classification. The major ones include mobility, resource management, security, IoT, low-latency and ultra-high bandwidth use cases. These different use cases can accommodate many industries, companies and end users that the 5G network will be serving; hence, the classification includes these use case subclasses. The major problem with some use cases, such as mobility, is that it must serve all the mobility aspects of the end user. In reality, mobility requires dynamicity and flexibility in the network slice that is serving the end user with the mobility requirement. Fortunately, network slicing facilitates the creation of a network with high flexibility.

Fig. 6