top of page

5G Standalone Network Architecture and Edge Computing use case

Updated: Apr 27

In the following article we will discuss the architecture and the main 5GC node and its functions, I have also included some potential use cases as requested by some of the readers.


Figure 1


5GC or standalone network functions


  1. Network Functions


  • Access and Mobility Management function or AMF


The AMF Represents the Control plane middle entity within the 5G network the AMF and holds the following responsibilities or functions :


Registration Management: Meaning registration or deregistration from 5G system it is a must for UE to complete registration in order to access to 5G services.

AMF must interact with other network functions through different interfaces to complete registration (Figure 3) .


Connection management: This is done through the N1 interface (Figure 1) this interface is transparent to NG-RAN ( like registration or authentification ) and connects UE to AMF and allows the exchange of the NAS signaling and moving UE to different connection states ( CM-Idle, CM-Connected ) .


Reachability management & Mobility management: Reachability management ensures that UE is always reachable and possible to page it in case a mobile terminated session is required to be established,this will move Ue to a CM-connected state and establish N1 interface .

Mobility management is used to maintain the knowledge of UE location within network.


Handling NGAP signaling: This represents the signaling between AMF and NG RAN through NGAP protocol 3GPP TS 38.413 this includes ( PDU session management, UE context management, paging procedure, transport of NAS messages, configuration transfert ..etc )



  • Session management function or SMF.


The SMF is a control plan function with the following responsibilities :


PDU session management: which represents a connection of UE to a specific UPF toward a specific Data network or DN as specified in Figure 2 require that the signaling with UPF to be established, the SMF use NAS to communicate with UE , the AMF and gNB manage the NGAP procedure but content are relayed to/from SMF when it is related to session management .


IP address allocation: A pdu session could be set to transfer either using IPV4 / IPV6

3gpp allows the UE to request the use of DHCP ( Dynamic host configuration protocol ) to

obtain an IP address and in this case, the SMF does not allocate an IP address but use the connectivity provided by the PDU session to obtain IP address from a DHCP server .

GTP-U tunnel management: GTP-U tunnels are used to transfer data between gNB and UPF

Downlink notification management. TEID ( allocated by UPF Via SMF & AMF ) will link the user plan packet to to a specific PDU session.


Downlink notification management: related to paging if data arrive within the UPF , the UE will need to be paged if data reaches the UPF once in Idle , the UPF will know that the UE does not have any active GTP-U tunnel toward gNB , so UPF will inform AMF through the SMF, which trigger the paging toward the UE.



  • User plane function or UPF.


Responsible of forwarding user plane packets between the gNB and UPF :

  • Uplink packets arriving from gNB use GTP-U tunnel to reach the UPF, the UPF will then remove the GTP-U header and forward the packet to the external DN ( Data network ).

  • Downlink packets arriving from DN will be mapped into a specific QOS flow belonging to a specific PDU session before transferring to the appropriate gNB.

  • The UPF also notifies the SMF in case downlink data arrives for UE in an Idle state ( during the paging procedure ).



What is QOS flow?


A QOS flow represents a set of packets having a similar QOS or lets say mapped to a specific QOS, but mapped how ? in fact the UPF will use the SDF template which is provided by the SMF and generated from the information provided by PCF or a service data flow to map a to a specific flow,

To understand how this works please think about it as a small map that includes ( source IP, Source port, destination IP, and destination port ) to map to a specific QOS flow, a PDU session can have multiple QOS flow see (Figure 2)



Figure 2


5G QOS flow mapping

Once the appropriate QOS flow is identified the UPF forwards the packet to the GTP-U tunnel belonging to the parent PDU session, there is always one GTP-U tunnel per PDU session ( refer to Figure 2 ) , where the QFI ( QOS flow identifier ) will be used to identify different QOS flow within a PDU session and will be included on GTP-U header.



  • Policy Control function or PCF.


The policy control function is responsible for providing policy relative to mobility management and policy associated with session management, session management is related to QOS provided by the PDU session.


for some, it may look confusing so let's draw how policy will be managed with all 3 entities. AMF , PCF , UDM ( Figure 3) :


  1. The AMF first obtains subscription information from UDM during the registration procedure

  2. The AMF then provide that information to PCF

  3. the PCF may modify the subscription information and setup certain predefined policy like restricting access to certain TAC in BH ..etc then PCF will modify the policy and send it to AMF

  4. Policy is then provided to UE (for example registration accept message )




Figure 3

Policy management for  AMF , PCF and UDM

  • Network slice selection function ( NSSF )


The major role of NSSF is to provide Network Slice Information to AMF, it is connected to AMF using the N22 interface, and the AMF can request the NSSF to complete the network slice function during the registration procedure.


But what does network slicing mean?


Network slicing refers to the allocation of a network resource to suit the requirement of a specific service and it includes resources from both core networks ( like UPF ) and RAN services like ( CU - UP) . For example under RAN a gNB can support multiple CU-UP each one under specific network slices.


A network slice is identified using the :

  • Slice /service type (SST) --> 8 bit which indicates an expected network slice behavior in term of feature and service ( 1= eMBB, 2 =URLCC, 3=mMTC)

  • A slice differentiator (SD) --> 24 bit is used to differentiate between network slices under the same SST .


Finally, We will have the AUSF (Authentication Server Function)  which Has been introduced in the 5G core (5GC) to manage the UE authentication, the Unified data management or UDM which provides a similar function to the HSS function on LTE the UDM will manage and store subscriber data, the UDR ( Unified data repository )& UDSF ( unstructured data storage function) which are responsible of storing UDM & other Network function respectively and without forgetting NRF which will allow the NF (network functions) to register their services and other NFs to recover it.



2. 5G Edge Computing use case


For a fan of baseball or UFC like myself we may know that for the people sitting in the back seat, the experience may be limited due to limited view or angles, something which can be enhanced using a 5G-powered VR Court-side, offering a more immersive and engaging way to enjoy the game however the combination ( throughput + latency ) may cause some reproduction challenges as most architectures are focussed on throughput or Latency but rarely optimized to deliver both at same time.


For example, a frame rate of 60 fps and a minimum (2K) resolution is required to avoid graphic pixelation where high throughput is required which is again increased in case of full omnidirectional representation ( higher viewport ) .


Also since the UE will deliver real-time sensor data in uplink where the rendering and produces the multimedia data needs to be performed by the XR server, a media which then needs to be delivered and scheduled to UE on downlink and displayed by the device So a low latency measured at transport or application layer may be also required.


To maximize the latency it is necessary to bring the computing capability closer to the edge ( locally on the event location ) or some nearby locations that could satisfy the latency budget ( Edge computing ) .

an example of such architecture is illustrated in Figure 4 , which assumes that the RAN ,UPF, and the LADN are co-located in the event location.


  • In our example, the PDU session will be configured to have multiple UPF ( Anchor UPF ) providing access to different data networks.

  • the first UPF will be localized ( for media content ) whereas the second UPF is assumed to be remote (for other content ) .

  • The SMF instructs the third UPF to act as a QUplink classifier ( UL CL ), the SMF provides the UPF with a set of traffic filters that can determine which packet should be routed to each Anchor UPF, ( the use of UL CL in this case is transparent to UE )


The LADN is a data network that provides service access across the specific local area and the UE can request LADN information during the registration request and provided by AMF during the registration response, the LADN will include a DNN & a list of one or more tracking areas.




Figure 4

5GS  Edge computing use case




Source used for this article & references :


3GPP TS 38.413

5G in bullet




















149 views0 comments

Comentários


bottom of page