ZSM-based Management and Orchestration of 3GPP Network Slicing: An Architectural Framework and Deployment Options

Driven by closed-loop automation, the Zero-Touch Network and Services Management (ZSM) framework offers invigorating features for the end-to-end management and orchestration of a sliced network. Although ZSM is considered a promising framework by 3GPP, there is a lack of concrete ZSM-based solutions for the 3GPP Network Slicing management. This article presents an architectural framework of a ZSM-based management system for 3GPP Network Slicing. The proposed framework employs recursive ZSM management domains to meet the specific management requirements of the standard 3GPP Network Slicing framework. Two deployment options are considered for the presented framework. The first option features the integration of the standard services of the 3GPP Management System within ZSM. The other one considers ZSM as a complementary system for the standard 3GPP Management System. From these deployment options, some key questions are identified on ZSM's interoperability with the existing 3GPP Network Slicing systems. Finally, for each option, the architectural and operational feasibility of the ZSM interoperation with 3GPP Network Slicing systems is shown through an example use case.


page 1

page 3


Predictive Closed-Loop Service Automation in O-RAN based Network Slicing

Network slicing provides introduces customized and agile network deploym...

Learnings Options End-to-End for Continuous Action Tasks

We present new results on learning temporally extended actions for conti...

A Decomposition and Metric-Based Evaluation Framework for Microservices

Migrating from monolithic systems into microservice is a very complex ta...

Responsibility Management through Responsibility Networks

The safety management is critically important in the workplace. Unfortun...

Model-Driven Process Enactment for NFV Systems with MAPLE

The Network Functions Virtualization (NFV) advent is making way for the ...

Big Memory Servers and Modern Approaches to Disk-Based Computation

The Big Memory solution is a new computing paradigm facilitated by commo...

Network Coding Function Virtualization

Network Functions Virtualization (NFV) and Network Coding (NC) have attr...


Management and orchestration of 3GPP Network Slicing brings complex challenges. Indeed, with the inherent flexibility of deployment offered by the Service-Based Architecture (SBA) [1], the standard 3GPP Network Slicing framework offers necessary features to support highly diverse services and use cases. In addition to multi-service and multi-tenancy – the hallmark attributes of network slicing – the real-world deployment of network slicing would exhibit multiple and distinct administrative and technological domains [2]. Supporting such diversity over the often shared physical infrastructure, imposes stringent requirements for the management and orchestration of network slices. The traditional management systems, which include the standard 3GPP Management System [3], are inadequate to meet such requirements.

Zero-Touch Network and Service Management (ZSM) framework, proposed by the European Telecommunications Standards Institute (ETSI), offers the desired capabilities for effective end-to-end (E2E) management and orchestration of complex network slicing environments. Its primary aim is to achieve ideally 100 percent automation of management and orchestration tasks by employing different automation techniques including Artificial Intelligence (AI) [4]. It has also been recognized among the key enablers for future 6G networks [5]. Due to the potential of ZSM, the 3GPP also supports delivering its standard management services through ZSM [3]. While a high-level framework for ZSM as a management framework for 3GPP Core Network (CN) and Radio Access Network (RAN) is given in [3], there is a lack of concrete ZSM-based solutions which can meet the specific management requirements of the standard 3GPP Network Slicing framework.

This article presents an architectural framework of ZSM-based management system for 3GPP Network Slicing. The proposed solution is based on recursive ZSM management domains [6]. Apart from inheriting the ZSM benefits, the presented framework can promise the specific management needs of the standard 3GPP Network Slicing framework. Specifically, the proposed architectural framework promises different degrees of management isolation for 3GPP network slices and enables scalability of the management system – both of which are recognized as the primary requirements of network slicing management systems [7].


Considering the commonly understood three-layered architecture for network slicing (comprising the management, network slice instance, and resource/infrastructure layers), we have used the following terminologies throughout this article. The term 3GPP Network Slicing systems generically refers to systems at either of the three layers. These systems include the standard 3GPP Management System at the management layer, and the standard 3GPP network slice instances at the network slice instance layer. At the resource/infrastructure layer, the ETSI-specified Network Functions Virtualization (NFV) systems are considered since these are the widely used systems in most network slicing architectures. The term standard 3GPP Network Slicing framework refers to the standard framework of network slice instances as specified by the 3GPP. However, the management of the standard 3GPP Network Slicing framework, inevitably involves the management and orchestration operations at the resource/infrastructure layer, in addition to the management operations of the network slice instances. The term 3GPP Network Slicing is used as an abbreviated form of the standard 3GPP Network Slicing framework, and is used interchangeably with the latter. Finally, the E2E 3GPP Slice refers to a 3GPP network slice instance comprising (potentially multiple) RAN, Transport Network (TN), and CN domain(s).


The rest of this article is organized as follows. The next section presents a brief overview of the recent advances in ZSM-based management and orchestration of network slicing. The proposed ZSM-based architectural framework for 3GPP Network Slicing management is presented in the subsequent section. Two deployment options of the proposed framework with the standard 3GPP Management System are then introduced. The first option features the integration of the standard 3GPP Management services into ZSM. The other option considers deploying ZSM as a complementary system with the current 3GPP Management System. This section also discusses some open questions concerning the ZSM’s interoperability with the existing systems in the 3GPP Network Slicing. The following section then highlights the architectural and operational interoperability of the proposed framework with the aforementioned 3GPP Network Slicing systems. An example, Virtualized Network Function (VNF) scaling process, is taken as a use case to describe the architectural and operational aspects in both deployment options. The final section concludes the article.

Related Work

Supporting multi-domain E2E management and orchestration through AI-based closed-loop automation is the cornerstone feature of ZSM. An AI-based cross-system management and orchestration solution in Reference [8] adopts Distributed Artificial Intelligence

techniques. Such techniques, which include transfer learning and federated learning etc., employ

intelligent agents in different management domains. Seamless collaboration between these agents enables the E2E network and service management across different domains. In Reference [9], collaboration among closed-loops in management domains is investigated to enable E2E service management in a multi-provider, multi-vendor, and multi-tenant environment.

Some solutions focus on enriching ZSM with additional features. For example, Reference [10] proposes a security enhancement to the ZSM architecture, addressing its various identified security threats. The work in [11] introduces the concept of Knowledge Plane (KP) in ZSM. As a pervasive system, the KP provides high-level models of the network’s overall functioning. Based on KP, other elements in the network receive services and advice about their role and operation. To deal with trust and reliability issues for entities which belong to different administrative domains, block-chain technology is adopted within the ZSM framework in [12].

Our Contribution

Despite the current progress, the adoption of ZSM for the standard 3GPP Network Slicing framework has not been sufficiently explored. Apart from its inherent capabilities, ZSM for the standard 3GPP Network Slicing framework is also required to meet its specific management requirements. More specifically, the capability to promise management isolation among slices is critical since the standard 3GPP Network Slicing framework supports NFs sharing among different slices, in addition to the dedicated NFs. Furthermore, the scalability of ZSM systems is also essential for such deployments since continual provisioning and management of massive slices is expected in real-world scenarios [1]. The next section presents our proposed architectural framework for ZSM-based management of 3GPP Network Slicing that can ensure these capabilities.

Fig. 1: The proposed architectural framework.

Proposed Architectural Framework

In this section, we present the proposed architectural framework of ZSM-based management system for 3GPP Network Slicing. The management isolation and system scalability demand a modular ZSM management system – a system exhibiting a high granularity of management domains, in addition to granularity in the management functions/services. As per the current ETSI ZSM specifications, such modularisation and granularity can be achieved by applying the ZSM architectural option of recursive management domains [6].

In this option, a ZSM Management Domain, or a set thereof, is composed of other Management Domains [6], over which an E2E service can be deployed. In essence, recursion enables the creation of multiple (possibly infinite) management domains, each with varying degrees of self-containment and self-dependence. Depending on the management and orchestration needs of the underlying domain, the management functionalities of different management domains can be similar or completely different from each other. As noted in [6], the option of recursive deployments is also applicable for the E2E Service Management Domain. Such deployments become useful for managing slices that may cover very wide coverage areas, and be potentially composed of several technological and administrative domains. Our proposed framework featuring the recursive ZSM management domains is shown in Figure 1.

In the proposed framework, the E2E service management domain, responsible for the E2E service management and orchestration, is composed of E2E RAN, E2E CN, and E2E TN service management domains. Such architecture allows the deployment and management of service instances independently over RAN, CN and TN domains, which have distinct technological and topological characteristics [13]. To focus on the interoperability of ZSM with 3GPP Network Slicing systems and also for simplicity, we only consider the CN domain for 3GPP Network Slicing henceforth. The control plane of the 3GPP CN domain constitutes SBA, and hence we use the terms SBA and CN control plane interchangeably.

The CN control plane NFs which can be optionally shared among different slices include Access and Mobility Management Function (AMF) and Network Repository Function (NRF). In comparison, the NFs dedicated to a slice (or slice-specific NFs) in any typical deployment include Session Management Function (SMF), Policy Control Function (PCF) and Network Data Analytics Function (NWDAF). In addition, a set of overarching NFs at CN, which are common to all slices, also exist which includes the Network Slice Selection Function (NSSF). The instances of other NFs, such as Unified Data Management (UDM), Unified Data Repository (UDR), and Unstructured Data Storage Function (UDSF), can be slice-specific, common, and overarching, depending on the deployment needs. Correspondingly, in the proposed framework, three levels of CN management domains are defined for slice-specific, shared, and overarching NFs of the slice, respectively.

The modularization of ZSM eases its programmability. It enables the provisioning of tailored, and slice-specific management domains when a request for slice/service creation is received from a ZSM Consumer (e.g., a vertical). A tailored management domain provides the slice owners (e.g., the verticals) with distinct options for management. It allows them to enforce their desired policies as well as control the management services exposure and data sharing towards the ZSM Cross-domain data services. Hence, despite sharing some NFs with other slices, management isolation to slice owners can be promised. In addition, a modularized ZSM system also facilitates the independent addition of new management domains at any level as new slices are added, thus making the ZSM system scalable. In essence, the modularized ZSM system, with management isolation and scalability attributes, becomes an enabler for autonomous management and orchestration operations in individual domains constituting the E2E 3GPP slice. Similar architecture designs based on recursive ZSM management domains may be considered for multi-domain RAN and TN architectures as well. They are, however, omitted here.

The NFV-based virtualization environments (e.g., NFVI-PoPs) have demonstrated suitability for deploying VNFs constituting a slice. For such deployments, apart from regular ZSM management domains, Virtualization management domains are also necessary. A Virtualization management domain for NFV employs the standard NFV Management and Orchestration (NFV-MANO) components as ZSM services. It is responsible for resource-level management of VNFs. That is, the compute, networking, and storage resources of, for instance, the slice-specific NFs (in Figure 1) will be managed by the Virtualization management domain. However, any application-level interaction of NFs such as NWDAF and PCF will still occur through the regular Management Domain services. Note that in contrast to multiple Virtualization management domains for different NFVI-PoPs shown in Figure 1, a single Virtualization management domain would suffice for deployments where all NFs of a slice can be hosted at a single location (e.g., alongside the Overarching NFs management domain).

Fig. 2: Deployment Options: Integrated (Option 1A and 1B), and Complementary (Option 2) (*Assuming NFV-based control plane NFs deployment only).

Deployment Options

The ongoing standardization efforts for ETSI ZSM are still at an early stage. Their specifications are preliminary, and are based on a higher level of abstraction. In this regard, envisioning the ZSM deployment scenarios in the real-world slicing environments is essential, as it can foster a practical way forward for identifying and addressing other challenges and providing their practical and viable solutions. Hence, having described its ability to promise management isolation and scalability, we now focus on the deployment options for the proposed framework.

The first deployment option features the integrated ZSM and 3GPP Management systems. This option is consistent with 3GPP’s conceptualization of ZSM system for 3GPP Network Slicing management [3]. Generally, such deployments will also incorporate an integrated NFV-MANO within ZSM [6] (Option 1A in Figure 2). However, standalone NFV-MANO deployments with the integrated ZSM and 3GPP Management System are also possible [14] (Option 1B in Figure 2). In Option 1A, the 3GPP Management services, and NFV-MANO functionalities become part of the ZSM services set. Hence, services of either system can be mutually invoked by authorized services of any other system. However, for Option 1B, the inter-operation of standalone NFV-MANO with the integrated ZSM and 3GPP Management System brings some interworking challenges, which will be highlighted later.

The higher modularization of ZSM management domains are expected to exhibit multiple closed-loops for automation at different levels of hierarchy. This would result in high computational complexity. Accordingly, its feasibility to provide a complete set of management services for large-scale and massive number of slices becomes challenging. In this regard, the operator may wish to utilize ZSM services selectively for chosen slices only. That is, the ZSM system may only act as a complementary system to the existing 3GPP Management System. In such deployments, ZSM, 3GPP Management System, and NFV-MANO will act as standalone systems interacting with each other through suitable interfaces. This is represented as Option 2 in Figure 2.

Open Questions

For each of the aforementioned deployment options, certain questions arise on ZSM’s interworking with the existing 3GPP Network Slicing systems. Firstly, some ZSM services represent an overlap with the standard 3GPP control plane NFs at the underlying slice instances. Secondly, interfacing challenges of ZSM arise with existing systems including the standalone 3GPP Management System, the standalone NFV-MANO and slice instances at the 3GPP CN control plane. We discuss these in the following.

Overlap of ZSM Services with SBA NFs

  1. Data Analytics: The NWDAF provides data analytics functionality at the SBA. This represents an overlap with the ZSM Data Analytics service. However, by distributing the data analytics functionalities among the control plane and the management plane, such overlap can be averted. As discussed in [15], such approaches can enable the deployment of AI-based algorithms for short-term forecasting at the control plane, while medium-term, and long-term forecasting at ZSM.

  2. Data Storage: The 3GPP specifications have defined control plane NFs such as UDSF, UDM and UDR, which are responsible for data collection and storage [1]. This also represents an overlap with ZSM Data Storage services. A ZSM system managing the slice requires constant, and up-to-date data retrieval from these NFs. Yet again, the data required for localized functionalities (e.g., data analytics at control plane) may only be stored at control plane resources, while the rest be transferred to the management plane entities.

  3. Policy Management: The ZSM Policy Management service and the 3GPP PCF are both defined for policy management related functionalities. With the interaction of control and management planes, it is possible to deploy static policies at the control plane through PCF. On the other hand, dynamic policies can be generated, for instance, through ZSM Domain Intelligence services, and then intermittently transferred to PCF for enforcement.

Interfacing ZSM with Existing Systems

  1. CN Control Plane NFs: For functionalities like data collection and other interactions, suitable interfacing between ZSM and SBA NFs is required. Since both ZSM and SBA are service-based, enhancing an SBA NF with Management Function interfaces as specified in [3] will enable such interaction.

  2. Interaction with 3GPP Management System: For Option 2 (Figure 2), where the 3GPP Management System exists as an external, standalone system, mutual services exposure from both systems is required for their interworking. The ZSM specifications have defined the ZSM Management Services Adapter for this purpose [6]. Similar functionality from 3GPP Management System side is performed by the Exposure Governance Management Function (EGMF). Since both ZSM and 3GPP Management System are service-based, such interaction can be realized with fewer challenges.

  3. Interaction between NFV-MANO and ZSM: The standard NFV-MANO interfaces are not service-based. This is a major challenge for ZSM’s interaction with the standalone NFV-MANO, as indicated in [14]. The ZSM Management Service Adapters to enable such interaction are required to translate the NFV-specific communication into ZSM and vice versa, and as such are expected to be more sophisticated.

  4. Other systems: At the resource/infrastructure layer, in addition to the VNFs, elements such as physical NFs and containerized NFs are also possible, which collectively constitute a slice. These elements will also require ZSM specific interfaces to interact with the ZSM System.

VNF Scaling Use Case

This section aims to delve deep into the technical aspects of the aforementioned deployment options, highlighting key services of the involved systems and their interworking. The process of VNF scaling is taken as a use case to describe both options. We choose this process since it can be triggered to meet both service management and the slice management requirements.

Integrated Option

We assume that the CN control plane NFs including PCF, UDM/UDR, UDSF, and NWDAF are enhanced with 3GPP Management Functions interfaces [3] as shown in Figure 3. The ZSM Domain Data Collection services can receive the continuous data streams from these NFs (1). The Domain Data Collection may choose to store the received data in the Domain Data Storage (2), or send it to the Domain Analytics (2). Based on the received data, and some additional information (e.g., of control plane topology and allocated resources to each NF), the Domain Analytics service can generate insights and predictions [6] within the purview of the current management domain. For instance, the Domain Data Analytics and within these the Anomaly detection service can detect anomalies in response time of a particular CN control plane NF at the slice-specific instance of the current slice (3). The Domain Data Analytics then invokes the Domain Intelligence services, which are responsible for deciding if any actions are required.

By employing AI techniques and running the closed loop automation, the Domain Intelligence services infer the need for VNF scaling to remedy the detected anomaly (4). Once the decision is made, the planning for action(s) is done. Such actions include Orchestration/Control actions, which are to be executed by the ZSM services from Domain Orchestration and Domain Control services. The Orchestration services, based on the available domain service model, which contains the complete description of infrastructure resources, can determine the feasibility and extent of VNF scaling (5).

The modification of resources at the resource/infrastructure layer is executed by the Domain Control services. Subsequent steps would normally involve the interaction of Domain Control services with the underlying infrastructure (6). However, in the integrated option, Option 1A, the Domain Control services are required to invoke the NSSMF services. For VNF scaling, the Resource Lifecycle Management service from Domain Control services invokes the Network Slice Subnet Provisioning Management service of the integrated NSSMF. This NSSMF service is defined in the 3GPP Technical Specification, TS 28.531 (6). If the NSSMF determines that the VNF scaling is possible, the NSSMF instance invokes the NFV Orchestrator (NFVO) services from the Virtualization management domain (7).

The NFVO will execute the VNF scaling (as per ETSI Group Specification, ETSI GS NFV-MAN001) as follows. After validating the received VNF Scaling Request, and performing the feasibility check, the NFVO sends the Scaling Request to the relevant VNF Manager (VNFM). The VNFM may perform certain preparation tasks (e.g., evaluating the request as per VNF lifecycle constraints) (8). Next, the VNFM invokes the Scale Resource operation of the NFVO (9). The NFVO sends a request to Virtualized Infrastructure Manager (VIM) to change resources (compute, storage, networking) required to scale the VNF. The VIM eventually modifies the resources to effectuate the resource scaling of the VNF (10).

Fig. 3: Integrated option (Option 1A) showing deployment of 3GPP Management and NFV-MANO services within ZSM.

Complementary Option

In the complementary option, the authorized 3GPP Management Services, for their specific operational needs, can invoke the ZSM services. For instance, the NSMF, through ZSM Management Services Adapter may invoke management services at the E2E Service management domain. The E2E Service management domain may in turn invoke the respective management domain’s services to start collecting data from the underlying slices and performing the analytics. Note that the actual slice provisioning and configuration in such deployments will still be carried out by the 3GPP Management System. That is, the ZSM system will not feature any Virtualization management domains. The ZSM overhead can also be reduced through the complementary option. For instance, the NSMF might subscribe to the Domain Analytics service to receive notifications for fault or anomaly detection etc. The NSMF may then instruct the NSSMF to perform the Fault Management operations to remedy the faults, thus relieving the ZSM system further from running the closed-loop automation.

For VNF scaling operation, following the steps (1) to (5) of the Integrated Option (in previous sub-section), the Domain Control services are required to invoke the external 3GPP Management System services (Figure 4). For VNF scaling, the NSMF’s Network Slice Provisioning Management Service exposed to the ZSM system can be invoked by the Domain Control services (6). The NSMF, in turn, invokes the Network Slice Subnet Provisioning Management Services of the respective NSSMF (7). NSSMF then utilizes the Os-Ma-nfvo interface of NFV-MANO [3] to send the VNF Scaling request to the NFVO (8). The VNF Scaling then takes place through VNFM and the VIM as described in the Integrated Option (in the previous sub-section) (9).

Fig. 4: Complementary option (Option 2) showing deployment of ZSM for 3GPP Management System.


ZSM is now considered an indispensable part of future management systems for network slicing. This article has investigated the applicability of ZSM in an operational 3GPP Network Slicing system. An architectural framework of ZSM based on recursive ZSM management domains is proposed. The deployment aspects of the proposed ZSM framework with 3GPP Network Slicing have also been discussed. Two possible options for deployment are discussed, one featuring the integrated 3GPP Management System within ZSM, and the other considering ZSM as a complementary system to the 3GPP Management System. For each option, we have shown the architectural and operational feasibility of ZSM interoperation with the existing 3GPP Network Slicing systems through an example use case.


  • [1] System Architecture for the 5G System; Stage 2 Release 16, 3GPP Std. TS 23.501, Rev. V16.9.0, June 2021.
  • [2] 5GPPP, “5G PPP Architecture Working Group - View on 5G Architecture (Version 3.0), Feb. 2020,” White Paper 2020.
  • [3] Management and orchestration; Architecture framework; Stage 2 Release 16, 3GPP Std. TS 28.533, Rev. V16.7.0, Mar. 2021.
  • [4] C. Benzaid and T. Taleb, “AI-Driven Zero Touch Network and Service Management in 5G and Beyond: Challenges and Research Directions,” IEEE Network, vol. 34, no. 2, pp. 186-194, 2020.
  • [5] W. Jiang, B. Han, M. A. Habibi, and H. D. Schotten, “The Road Towards 6G: A Comprehensive Survey,” IEEE Open Journal of the Communications Society, vol. 2, pp. 334-366, 2021
  • [6] Zero-touch Network and Service Management (ZSM); Reference Architecture, ETSI GR ZSM 002 V1.1.1 (2019-08), 2019.
  • [7] F. Z. Yousaf, V. Sciancalepore, M. Liebsch, and X. Costa-Perez, “MANOaaS: A Multi-Tenant NFV MANO for 5G Network Slices,” IEEE Communications Magazine, vol. 57, no. 5, pp. 103-109, 2019.
  • [8] M. Bagaa, T. Taleb, J. Riekki, and J. Song, “Collaborative Cross System AI: Toward 5G System and Beyond,” IEEE Network, pp. 1-9, 2021.
  • [9] M. Xie, P. H. Gomes, J. Harmatos, and J. Ordonez-Lucena, “Collaborated Closed Loops for Autonomous End-to-End Service Management in 5G,” in 2020 IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN), 2020, pp. 64-70.
  • [10] C. Benzaid and T. Taleb, “ZSM Security: Threat Surface and Best Practices,” IEEE Network, vol. 34, no. 3, pp. 124-133, 2020.
  • [11] F. Rezazadeh, H. Chergui, L. Christofi, and C. Verikoukis, “Actor-Critic-Based Learning for Zero-touch Joint Resource and Energy Control in Network Slicing,” in ICC 2021 - IEEE International Conference on Communications, 2021, pp. 1-6.
  • [12] V. Theodorou et al., “Blockchain-based Zero Touch Service Assurance in Cross-domain Network Slicing,” in 2021 Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit), 2021, pp. 395-400.
  • [13] M. Chahbar, G. Diaz, A. Dandoush, C. Cérin, and K. Ghoumid, “A Comprehensive Survey on the E2E 5G Network Slicing Model,” IEEE Transactions on Network and Service Management, vol. 18, no. 1, pp. 49-62, 2021.
  • [14] Zero-touch Network and Service Management (ZSM); Landscape, ETSI GR ZSM 004 V1.1.1 (2020-03), 2020.
  • [15] D. Bega, M. Gramaglia, R. Perez, M. Fiore, A. Banchs, and X. Costa-Pérez, “AI-Based Autonomous Control, Management, and Orchestration in 5G: From Standards to Algorithms,” IEEE Network, vol. 34, no. 6, pp. 14-20, 2020.