The architecture of the 5G Core Network. What to do: 5G Non-Stand Alone (5G NSA) or 5G Stand Alone (5G SA)?


First commercial 5G networks are launched everywhere, Europe USA, Korea. Recently many operators announced the commercial start of 5G networks, some of them, like Swisscom in Switzerland, even promised to deliver countrywide coverage till the end of 2019. Things are happening, but what about 5G SA?

Based on these announcements I asked myself the question. Does it mean 5G technology if fully capable, does it delivers all that was promised by standards?

To be first, to deliver first base station, cover some city, deliver first modem, first phone, is it just a pure marketing buzz word to show who is defining the leader board or real winner in the race?

To find the answers to my questions, we shall understand what are the services which were building blocks of 5G standard and how it differs from 4G or 3G networks.

ITU-R and ETSI/3GPP shall help to understand it better.

ITU-R’s definition of 5G mobile network use cases

Mobile data consumption of internet is growing year to year, mainly due to video streaming services, an increase of penetration of smartphones and an increasing number of parallel connections. At home, we have more devices requiring internet connections like mobiles, broadband modes, refrigerators, washing machines, heating systems, security systems etc.

The new 5G mobile technology assumes that it shall extend usability and quality for new use cases and applications, which can not be delivered today over 4G/LTE technology.

ITU-R has defined three main usage scenarios [ITU-R M.2083]:

  • Enhanced Mobile Broadband (eMBB) to deal with hugely increased data rates, high user density and very high traffic capacity for hotspot scenarios as well as seamless coverage and high mobility scenarios with still improved used data rates
  • Massive Machine-type Communications (mMTC) for the IoT, requiring low power consumption and low data rates for very large numbers of connected devices
  • Ultra-reliable and Low Latency Communications (URLLC) to cater for safety-critical and mission-critical applications

which requires different key capabilities as shown on Figure 1.:

5G SA - Figure 1. Key capabilities of ITU for 5G network
Figure 1. Key capabilities of ITU for 5G network

1. eMBB (enhanced Mobile BroadBand) is a usage scenario, which is going to be used by broadband services, live media. It requires high data peak ratio, wide are coverage, high efficiency in spectrum utilization (capacity per km2) good user experience data rate and of course low energy consumption on the end-user terminal. Typical solutions: video streaming services, online gaming (EDGE computing), Augmented Reality/Virtual Reality, broadband services. Latency for that segment shall fulfil the requirement of 10 ms for User Plane and 5-10ms for Control Plane. It is in practice, improved service of already existing 4G/LTE broadband access.

2. mMTC (massive Machine Type Communication) usage scenario, designed for IoT solution, requires the support of millions of devices communicating in parallel but with smaller data peak rates, relatively low energy consumption and limited mobility. Smart city, smart Home are typical use scenarios for that case. Latency for that segment shall fulfil the requirement of 10 ms for User Plane and 5-10ms for Control Plane.

3. URLLC (Ultra-reliable Low Latency Communication) use cases, are very specific for safety-critical scenarios and require extremely low latency and very high reliability and mobility. The typical customers will be self-driving cars, industry/production automation (Industry 4.0). Latency for that segment shall fulfil the requirement of below 5 ms, best 1 ms for User Plane!

Looking especially on low latency demand, we can see, that both 5G System elements: 5G NR and 5G NGC must be able to support it technologically. 3rd critical element is the ISP performance. To deliver all these services, the challenge is not only in the design and delivery of New Radio technology (new frequencies, new modulation technique, multi-array antennas, MIMO, etc.) but also design and implementation of 5G New Generation Core architecture 5G (NGC), able to support specially URLLC and mMTC low latency use cases.

3GPP in Q1 2019 closed standards Release 15, of new core network architecture. Let’s see what are the changes and challenges.

3GPP approach to new 5G core network architecture (5G NSA and 5G SA)

It is gonna happen for the first time in mobile industry, that the core network and ISP connectivity will become a bottleneck in the network latency case when new 5G NG radio solution will be very efficient in latency delivery. Based on IUT-R assumptions, clearly we can see it requires a change in the Core network architecture to support case mMTC and URLCC.

3GPP, following network operators suggestions, defined the 5G core network standard based on three different phases:

  • early Rel-15 drop” so-called Non-standalone NR (NSA NR) which allows the connection of NR base station (called gNB) to existing Evolved Packet Core network (EPC).
  • “regular Rel-15 freeze” so-called standalone NR (SA NR), competently new core network architecture with new functions and interfaces, intended to support connection only of gNBs.
  • late Rel-15 drop” where LTE base station will be connected to NR SA via gNB (Control plane and user plane – architecture in option 4 or User plane directly to NGC and control plane over gNB – architecture in option 4a) or vice versa (option 7 and 7a) [3GPP TS38.801]. This version was added on the last stage of standardization, to let utilize LTE network as well in SA case (multi-access).

The first phase, “early Rel-15 drop” was designed to help launch first 5G networks already in 2019-2020, based on existing core network nodes. It requires mainly software modifications into existing core network elements. This drop was in that case answer to the first part of the question. It is an early phase solution, not ready to support all the ITU-R cases.

“regular Rl-15 drop” defining new architecture SA NR standards was approved by 3GPP Release 15 just in Q1 2019. This standard is delivering a completely new solution into the operator’s garage, but not the final one, as 3GPP Release 16 will be defining new elements of service function to support URLLC products.

Now, in 2019 we have a dilemma. Connect 5G NR to the existing core network, or wait and change the core network into NGC SA? What makes sense?

In my opinion, the support to the answer is in the definition of use cases from ITU-R mentioned previously. If we think only about the implementation of eMBB, at least in the very basic scenario, we could “refresh” existing core network and it would be enough to start.

However, if the ambition is to implement also mMTC and URLLC use cases, it will demand to step into new 5G SA solution, and in my opinion, it makes sense from various reasons, as soon as possible to do it.

The new low latency requirement, demand from mobile operators new architecture, not to forget about better ISP providers. Both elements of the communication chain become crucial.

What are the changes to the architecture and technology basics of the new core network?

5G SA – new generation of Core Network

I will describe the benefits of new SA NR architecture as well as new technological assumption using the 3GPP standard. 

5G SA basic architecture
5G SA basic architecture

General assumptions of 5G SA are described in document [3GPP TS23.501]:

The 5G System architecture is defined to support data connectivity and services enabling deployments to use techniques such as e.g. Network Function Virtualization and Software Defined Networking. The 5G System architecture shall leverage service-based interactions between Control Plane (CP) Network Functions. Some key principles and concept are to:

  1. Separate the User Plane (UP) functions from the Control Plane (CP) functions, allowing independent scalability, evolution and flexible deployments e.g. centralized location or distributed (remote) location.
  2. Modularize the function design, e.g. to enable flexible and efficient network slicing.
  3. Wherever applicable, define procedures (i.e. the set of interactions between network functions) as services, so that their re-use is possible.
  4. Enable each Network Function to interact with other NF directly if required. The architecture does not preclude the use of an intermediate function to help route Control Plane messages (e.g. like a DRA).
  5. Minimize dependencies between the Access Network (AN) and the Core Network (CN). The architecture is defined with a converged core network with a common AN – CN interface which integrates different Access Types e.g. 3GPP access and non-3GPP access.
  6. Support a unified authentication framework.
  7. Support “stateless” NFs, where the “compute” resource is decoupled from the “storage” resource.
  8. Support capability exposure.
  9. Support concurrent access to local and centralized services. To support low latency services and access to local data networks, UP functions can be deployed close to the Access Network.
  10. Support roaming with both Home routed traffic as well as Local breakout traffic in the visited PLMN.

We can add also, that the 3GPP decided to change the standard of signalling for control plane into HTTP/2 over TCP/IP and exchange of information composed of JSON payload (11). It is another BIG fundamental change, giving telecom vendors and operators access to wisely used for years: tools, software libraries, and programming solutions from the IT software development market.

It is a revolutionary change from “closed, strictly defined core network nodes/elements architecture” towards modern software-defined architecture utilizing service-based applications communicating each other (not always, however) over RESTfull interfaces. It allows building multi-layer clouds typically dividing it into central, regional up to local cloud system (EDGE computing in 3GPP called Multi-access Edge Computing – MEC).

What are the practical arguments to each of assumption (numbers refer to the previous counting):

1. Separation of UP and CP let the operator implement multi-layer infrastructure, which will be critical to delivering low latency solution for mMTC and URLLC cases. As I already mentioned, the core network must be much faster than today delivered solutions from mobile operators and ISP providers. It will be possible only in case of pushing the Network Functions and applications into the EDGE of the network, work locally, deliver locally. It does not mean only access to the internet, as eMBB is a separate issue. If we understand that URLLC cases like Industry 4.0 or V2X between vehicle communication require control over fast-responding applications, we need to put the applications as close to the geographical location of the customer, as possible. Examples: City infrastructure (V2I) to aim to communicate with all vehicles (V2V) and pedestrians (V2P) will require more than one local MEC system, due to its size.

We can imagine few concentrated Virtual RAN systems covering parts of the city, each requiring separate MEC support. The movement of the vehicle in the city, changing the Virtual RAN zones requires inter MEC communication and here the separation of UP and CP delivers support.

Multi level UPF network structure of 5G SA network

2. Network slicing is a concept, which allows separating “physical” resources of radio and core network for separate service delivery applications. As long as the mobile network operator will deliver locally enough radio capacity, we will be able to give higher warranty of service to separate slices, with different QoS conditions, specific for end-user needs – V2X vehicles control, control of production line, etc.. Network slicing will let operators give different priorities for other services (Standardized SST values: eMBB -1, URLLC – 2, MIoT -3). Another interesting feature of slicing is the possibility to implement it regionally, e.g. deliver low latency service locally to the industrial zone, stadium. It simplifies launch strategy and minimizes associated costs for the operator. Slicing, 5G SA based, is crucial for mMTC and URLCC use cases.

Functions/applications will have an impact on the New Radio parameters including QoS. It means it would be possible to influence Quality of Experience (QoE) for some of the users from external applications, like the MEC system.

3. Defined procedures as a service will let operators faster and easier develop new functionalities base on already set up communication procedures between functions and applications.

4. Direct interaction between network functions (NF) is a revolution in the mobile telecommunication standards. In new concept, relevant NFs will be able to communicate directly with other NF without a need to push the communication over one central point (today it is a case due to continuity of SS7 signalling network architecture). The exception is defined for Non-3GPP (LAN based) systems which must be interconnected over N3IWF function. So, there won’t be network elements/nodes, but only network functions, which may be elastically realized on different virtual systems. Next benefit is a simplification of the communication interface between NFs, it will be realized over service-based interface bus and communication with external world will utilize API based on OpenAPI 3.0.

5. Interface communication between Access Network (AN) or none-3GPP access and Core Network (NGC) is going to be standardized.

6. Unified authentication framework, allows the development of new services using simplified service procedures (OAuth 2.0).

7. Stateless NFs, is another technology required to support low latency solution, but also to adopt to modern REST-based coding. This is one of the most fundamental change, as any synchronize communication between Control Plane (CP) and User Plane (UP) was negatively influencing the speed of internet and increasing latency.

8. Capability exposure will simplify the development of new services. Thanks to this capability, 5G Core Network and Application Function will be providing information to each other via NEF function or directly.

9. UP functions deployed close to the Access Network in parallel to a central location, is another benefit of spiting User Plane system (UPF). It let influence lower latency and speed of delivery of the internet.

10. Support for Home Route roaming as well as LBO connection, are crucial for fast and world-wise solution. Looking at the difficulties of roaming VoLTE roll out not standardized from the beginning between operators, we can see low penetration of that technology today. 5G roaming scenario from the beginning is standardized and hopefully will be seriously implemented by all the industry.

11. Last, but not least is the development of future communication protocol. In 3GPP Rel 15 standard, it was decided to change it into HTTP/2 over TCP/IP with TLS as a security layer. It is not an optimal solution, as TCP/IP is not supporting asynchronous communication, and has many bottlenecks in terms of supporting fast communication from first seconds (start-up Window). ETSI and 3GPP are very aware of these problems. The next change followed in 3GPP Rel. 16 is going to change the communication protocol into HTTP/2 over QUIC and UDP/IP. Benefits of QUIC are 0ms latency for known connection (perfect for 5G network) as well as inbuilt security, which does not require TLS, again impacting positively on decreasing latency and increasing speed.

ETSI is working also on New Generation Protocol for future (NGP protocol), but results of it we shall expect above 2025. My prediction is that change to HTTP/2 in CP is going to happen as soon as first 5G SA implementations will appear. Today User Plane (UP) connections are dominated by TCP/IP and IPv4. Let’s hope it is not going to follow a very bad history of IPv6, and UP connections will smoothly transfer int QUIC/UDP and IPv6 technology.

Ease of implementation – “Beauty and the Beast”

Merging two worlds, DevOps methods on cloud base systems with telecom regime shall give very interesting results. Many software developers, ready to use microservices based on containers, can deliver NFs in a relatively short time. Standardization of communication and definition of NFs functions is a good receipt for product delivery.

The new 5G standard SBA based, controversially is just as easy for development for IT software houses like for “BIG” telecommunication vendors (Affirmed Networks, Ericsson, Huawei, Nokia, ZTE, etc.). We can also recognize the development of small firms, announcing first versions of virtual Radio (Open Radio Access Network (O-RAN Alliance) or virtual EPC (OMEC Open Mobile Evolved Core, from Open Networking Foundation (ONF)). Soon we shall recognize more and more competition on that field.


Thinking out loud, the 5G NGC architecture is the must, but. Any new 5G network soon commercializing will be based mainly on 5G NSA architecture, as it is simpler and faster to deliver the eMBB case. Implementation of additional 5G radio units (RRU) on existing locations, and start delivering a simple broadband solution with higher speed and relatively lower latency is simple, the operational effort for operator and vendor.

The 5G SA implementation today may look like difficult due to lack of terminal support – Qualcomm’s modem “X50” do not support yet 5G SA core network protocols, the Qualcomm “X55” supporting higher speed and 5G SA is planned for the end of the year 2019. Huawei’s “Balong 5000” modem is supporting SA according to specification from the beginning but will be launched the second half of the year 2019. Analyzing that aspect from a time perspective, implementation of 5G SA architecture will take about a year, so the massive deliveries of 5G SA based terminals will meet this time frame, too.

The business case shows that it would be a mistake, not to start now in parallel to 5G NSA, study and implement of first 5G SA systems.


  1. ITU-R M.2083
  2. 3GPP – TS23.501
  3. ETSI – NGP principles
  4. O-RAN Alliance
  5. ONF (Open Networking Foundation)
  6. Qualcomm
  • If you like the text – share it with your friends.
  • Please leave a comment. It is a source of guidance for me.
  • You can follow me on LinkedIn and Facebook.