5G/NR - Network Slicing

zsyaaron

Network Slicing

Data between UE and Network (or another UE) go through various components on the data path. In most case, the resource allocation and the data path is configured statically or semi-statically. We cannot say all those components are optimized for each individual user or each individual use case (e.g, rush time traffic, regular hour traffic, eMBB, IoT etc). However, in ideal case where you can configure resource allocation and parameters of the components along the data path dynamically (by automation), we may define a set of parameters of all the components on the data path in most optimal way for specific UEs or specific use cases. The specific set of parameters assigned for the UEs or use cases is called a 'Slice' of the network. Network Slice is a logical concept of splitting all the resources along the data path into multiple sets, each of which is optimized for specific UEs or use cases.

Followings are the list of the topics we want to look into

• Implementation Requirement
• Slicing vs 5QI(QoS) vs DNN(APN)
• Network Slicing vs Private Network
• Signaling for Network Slicing
• Who decides which slide to use ?
• Identification of Network Slice
• So many different types of NSSAIs
• Get the Test Procedure and Log / Amarisoft TechAcademy
  

Implementation Requirement

I don't think there is any specification which explicitely defines all the details on how to implement these slices. But by the nature of dynamic configuration changes, we would easily guess that all the network elements along the data path should be constructed in such a way to change the configuration by remote control and automation. That's why the concept of network slicing can easily be achievable in 5G where most of the network components are designed for virtualization and the components along the transport (e.g, routers) is designed as SDN (Software Defined Network).

Slicing vs 5QI(QoS) vs DNN(APN)

Since Network Slicing is a concept of configuraing all the component along the end to end communication path, it can do everything that QoS(5QI in case of NR) and DNN(APN). That is, a certain aspect of network slicing and QoS and API overlaps. Due to this, we are often confused with these three concepts.  

In short, I would say Network Slicing would be a superset of QoS and APN. That is, Network Slicing can do everything that QoS and APN do, but not in vice versa, meaning that there are certain aspect of Network Slicing that cannot be done by QoS or APN.

Network Slicing vs APN : Let's first think of Network Slicing and APN. Basically APN is controlling network parameters (or configurations) by switching the traffic to various different gateway on the core network. We can assign different APN(gateway) depending on MNO (Mobile Network Operator) or MVNO (Mobile Virtual Network Operator). In other way, we can assign different APN(gateway) depending on the service types (e.g, internet, mms, sms, voice etc). We can do all of these with Network Slicing as well. As you may notice here, APN has control only over the components after the gateway, but it cannot control over the component before the gateway. For example, APN does not have any control over radio access network, whereas Network Slicing can controll the configurations not only for APN path but also for Radio Access path as well.

Network Slicing vs QoS : Now let's compare Network Slicing with QoS. Difference between Network Slicing and QoS is more subtle since QoS is also controlling the data flow end-to-end to meet a certain SLA(Service Level Agreement). The main difference between QoS and network slicing is that QoS focuses on controlling the performance and service levels that a network can provide to different traffic types and applications, while network slicing enables the creation of separate virtualized sub-networks that can be tailored to meet the specific requirements of different applications. While QoS can be used to define and enforce specific performance requirements, network slicing provides a more flexible and granular approach to service provisioning, enabling the creation of customized sub-networks that can meet the specific needs of different applications. Followings are some of the examples showing these differences in terms of use case implementation.

• Streaming video: A streaming video service requires a high data rate and low latency to deliver high-quality video content. In this case, QoS can be used to prioritize video traffic and allocate more network resources to ensure a smooth and uninterrupted viewing experience for users. Network slicing, on the other hand, can be used to create a separate slice for the video service, with customized QoS parameters and resource allocation, ensuring that the video service has the necessary resources to deliver high-quality video content.
• VoNR : QoS and network slicing are both used to support VoNR services in 5G networks, but they serve different purposes. QoS is used to define and enforce specific performance requirements for VoNR traffic, such as low latency and low packet loss. This is critical for ensuring that VoNR services provide a high-quality user experience. Network slicing, on the other hand, can be used to create separate slices for different VoNR services with customized QoS parameters and resource allocation. This enables the network to provide differentiated voice services that are tailored to meet the specific needs of different users and applications. For example, a network slice can be created for emergency services with high priority and guaranteed resources, ensuring that emergency voice calls are always delivered with the highest quality and reliability.
• Industrial automation: Industrial automation requires ultra-low latency and high reliability to support real-time control of machines and systems. QoS can be used to provide the required level of performance and service for industrial automation traffic, but it may not be sufficient to guarantee the necessary level of reliability. In this case, network slicing can be used to create a separate slice for industrial automation, with customized QoS parameters, resource allocation, and security policies, ensuring that the industrial automation traffic has the necessary level of reliability and security.
• Autonomous vehicles: Autonomous vehicles require low latency and high bandwidth to support real-time communication and decision-making. QoS can be used to allocate more network resources to the autonomous vehicle traffic, but it may not be sufficient to provide the necessary level of performance and service. In this case, network slicing can be used to create a separate slice for the autonomous vehicle service, with customized QoS parameters and resource allocation, ensuring that the autonomous vehicle traffic has the necessary level of performance and service.
  

But there are still some of the network slicing feature that cannot be done by QoS. For example, both Network Slicing and QoS can differentiate SLA between different application type (e.g, SLA between VoNR and eMBB), but QoS cannot differentiate the SLA within the same application. For example, QoS cannot differentiate the SLA between one UE (subscriber, also called a tenant) for the same application (e.g, VoNR) whereas Network Slicing can do.

Network Slicing vs Private Network

As you may notice from previous sections, there are many different ways of slicing the network. Let's suppose the Opeartor Y sliced network as illustrated below. In this example, we see a slice (i.e, dedicated network) for BMW group and FIAT group. From the point of BMW group and FIAT, it would be as if they have their own private network. That is, Network Slicing can be used as a way of implementing Private Network in wider sense of meaning. You may not say this cannot be a private network since the Operator Y would use licensed spectrum. I also agree that a keyword for Private Network was 'Unlicensed Spectrum', but in many case the meaning of technical terms varies (expands in most case) and I think the meaning of Private Network also has evolved to encompass various use case like this. I think this type of Private Network would attract attention not only from private company but also from Network Operators. To be hones, I don't think the private network based on Unlicensed Spectrum would be a strong motivation to existing operators, but this type of private network would be attactive to the established network operators as well.

Source : 5G network slicing: automation, assurance and optimization of 5G transport slices

Signaling for Network Slicing

Network slicing signaling process happenes at a few different stages like Initial Attach, PDU Establishment, Policy Change. Probably the most important step would be the process at Initial Attach. PDU Establishment would be mostly for defining various QoS Flow and Polich Change would be to associate the specific slices to a specific UE/Application as specified in the Policy Rules.

Following is the brief signaling flow at initial attach related to network slicing.

• When UE send Registration Request it specifies NSSAI. 3GPP defines a few large group of slice service/type as described here. By this NSSAI in Registration Request message UE tells Network saying that I want to get access to this and this type of slice.
• Then this requested Slice Information is tranferred to UDM. UDM check if the requested slice is allowed for the specific UE. If it is allowed, the UDP accept the request. If not, it rejects the request.
• Once the slice request is accepted, (with a few more steps of checkups on core network side) the acceptance is notified to UE with a few additional information like Configured NSSI, NSSI Inclusion Mode etc.
  

Source : Recreated from MAS5G: Move Around Smartly in 5G (IEEE)

Followings are the IE (Information Elements) involved in the signaling shown above.

• Registration Request.Network Slicing Indication
• Registration Accept.Allowed NSSAI
• Registration Accept.Network Slicing Indication
• Registration Accept.NSSI inclusion mode
  

The procedure explained above is just about the request and accept process for a specific slice. It does not really create any data pipe (slice). Setting up the real data pipe and associating the pipe with a slice happens at later step using the following procedure.

Step	Direction	Message	Type of NSSAI
1	UE -> NW	Registration Request	Requested NSSAI
2	UE <- NW	Registration Accept	Allowed NSSAI
3	UE -> NW	PDU Session Establishment Request	Selected NSSAI
4	UE <- NW	PDU Session Establishment Accept	Selected NSSAI
5	UE <- NW	Manage UE Policy Command
6	UE -> NW	Manage UE Policy Complete

Who decides which slide to use ?

Of course, the final decision is done by network since network functions as a master in almost every communication, but the network slice selection can be triggered by either UE or Network or Both.  It depends on a variety of factors such as the type of service or application, the location and capabilities of the mobile phone, and the network conditions.

The mobile phone will typically have a UICC that stores the credentials of the slice it is authorized to use. When the phone connects to the network, it will use the credentials to authenticate with the network and request access to a specific slice.

The network can also decide which slice to assign to the mobile phone based on the type of service or application the phone is trying to access. For example, if the phone is trying to access a high-bandwidth video streaming service, the network may assign it to a slice that has a higher capacity for handling video traffic.

Additionally, Network can also use the mobile phone's location, device capabilities, and other information to decide which slice to assign to it. For example, if the mobile phone is located in an area with a high density of users, the network may assign it to a slice with a lower capacity to ensure fair usage of resources.

This signaling procedure can be illustrated in other way as well illustrated by Rahim Navael as shown below. Rahim kindly approved me to share the illustration in my note.

Image Source : Rahim Navaei's linked in post

High level description of the Network Slice selection process can be described as follows. The details are specified in 3GPP 23.501-5.15.5.2.1

• The UE initiates registration with the 5G network by sending a Registration Request message to the 5G Core Network (5GC).
• The Registration Request message includes a list of requested NSSAIs that the UE wants to use for the current registration.
• The 5GC receives the Registration Request message and checks the requested NSSAIs against the allowed NSSAIs configured for the 5G network slice.
• If the requested NSSAIs are allowed by the network, the 5GC selects a network slice based on the configured NSSAIs and sends a Registration Accept message to the UE.
• The Registration Accept message includes the allowed NSSAI and other information necessary for the UE to access the network slice.
• The UE receives the Registration Accept message.
• The 5GC keeps track of the configured NSSAI for each network slice in its database. When a UE initiates a registration, the 5GC uses this information to select the appropriate network slice based on the requested NSSAI.
• When UE want to establish a data session, it specifies a selected NSSAI in UL NAS Transport during PDU session establishement procedure. --> You may not specifies the selected NSSAI at this step, network would assign a default NSSAI as the selected NSSAI.
• If Network can use the selected NSSAI from UE, it would send PDU session establishment accept message with S-NSSAI which indicates the Selected NSSAI.
  

NOTE : For the details of requested NSSAI, allowed NSSAI and selected NSSAI, refer to this section.

As illustrated above, most of the important configuration of Network Slice is done in core network side, meaning you need to get the detailed understanding on Signaling for the related corenetwork interface. You may refer to followings notes for corenetwork details.

• N2 interface
• N8 interface
• N22 interface
  

Identification of Network Slice

The identification of a Network Slice is indicated by a NAS Information Element called S-NSSAI(Single Network Slice Selection Assistance Information) as structured as below. S-NSSAI is made up of two field SST (Slice/Service Type) and SD (Service Differentiator). SD is an optional field. SST has 8 bit field length implying that it can indicates a total of 255 different slice types.

The SST field may have standardized and non-standardized values. Values 0 to 127 belong to the standardized SST

range. Values 128 to 255 belong to the Operator-specific range.

NOTE : 23.003-28.4.2 states 'The SD field has a reserved value "no SD value associated with the SST" defined as hexadecimal FFFFFF. In certain protocols, the SD field is not included to indicate that no SD value is associated with the SST'.

< 23.003 - Figure 28.4.2-1: Structure of S-NSSAI >

• SST indicates Slice Service Type which is a mandatory item.
• SD indicates a kind of slice ID within the SST which is an option item. This is used to indicates a specific slices in case where there are one or more slices are supported in the same SST.
  

The SST standardized in 3GPP as of now are as shown. Only 5 (as of  23.501 version 17.5.0) out of 127 possible types that can be configured by SST field.

< 23.501 v17.5 - Table 5.15.2.2-1 - Standardised SST values >