PaperlayerView exact revision on arXiv

Source-linked AI summary

5G network slicing using SDN and NFV- A survey of taxonomy, architectures and future challenges

Alcardo Alex Barakabitze, Arslan Ahmad, Rashid Mijumbi, Andrew Hines

arXiv:1912.02802v1cs.NIcs.DCcs.MMeess.SY

TL;DR

The paper addresses the need to meet diverse 5G vertical requirements through network slicing and softwarized infrastructure. It surveys SDN/NFV-based technologies, architectures, projects, standards, orchestrators, and management approaches, concluding that holistic adaptive orchestration and RAN virtualization remain important challenges.

  • Problem

    5G network slicing must satisfy different verticals and tenants, but existing surveys provide limited coverage of standardization, industrial efforts, architectures, orchestration, challenges, and research directions.

  • Method

    The paper conducts a comprehensive survey of 5G network slicing using SDN and NFV across enabling technologies, architectures, initiatives, standards, orchestration, and deployment domains.

  • Results

    The survey compares architectural approaches, summarizes open-source orchestrators and proof-of-concepts, and reviews single-domain, multidomain, edge, fog, and RAN slicing management approaches.

  • Takeaways & Limitations

    Effective 5G slicing requires coordinated SDN/NFV softwarization, multi-domain orchestration, tenant support, and management spanning network, edge, cloud, fog, and RAN resources.

  • Takeaways & Limitations

    Holistic orchestration must still ensure slice service and ELA/SLA requirements while using underlying resources efficiently, requiring adaptive decisions based on current and predicted states.

Abstract

from arXiv · show

In this paper, we provide a comprehensive review and updated solutions related to 5G network slicing using SDN and NFV. Firstly, we present 5G service quality and business requirements followed by a description of 5G network softwarization and slicing paradigms including essential concepts, history and different use cases. Secondly, we provide a tutorial of 5G network slicing technology enablers including SDN, NFV, MEC, cloud/Fog computing, network hypervisors, virtual machines & containers. Thidly, we comprehensively survey different industrial initiatives and projects that are pushing forward the adoption of SDN and NFV in accelerating 5G network slicing. A comparison of various 5G architectural approaches in terms of practical implementations, technology adoptions and deployment strategies is presented. Moreover, we provide a discussion on various open source orchestrators and proof of concepts representing industrial contribution. The work also investigates the standardization efforts in 5G networks regarding network slicing and softwarization. Additionally, the article presents the management and orchestration of network slices in a single domain followed by a comprehensive survey of management and orchestration approaches in 5G network slicing across multiple domains while supporting multiple tenants. Furthermore, we highlight the future challenges and research directions regarding network softwarization and slicing using SDN and NFV in 5G networks.

1. Introduction

The paper frames 5G network slicing as a response to demanding service and business requirements across diverse verticals. It surveys SDN/NFV-based slicing concepts, architectures, initiatives, standardization, orchestration, and future challenges.

  • Motivation: 5G must address rapidly growing mobile video, IoT, and bandwidth-intensive applications while supporting substantially greater capacity and performance than 4G.The paper cites 1,000 times current 4G capacity, 10–100 times higher user data rates, 99.999% reliability, and sub-1 ms latency as expected targets.
  • Motivation: Network slicing can provide differentiated services for enterprise verticals by enabling operators to customize network operations.The paper links network slicing to a projected $300 billion enterprise opportunity by 2025.
  • Research gap: Earlier surveys addressed selected aspects of network slicing, but recent work is limited in review breadth, standardization coverage, industrial projects, challenges, and research directions.The paper specifically identifies incomplete coverage of SDN, NFV, cloud/edge computing, and intelligent techniques.
  • Scope and contributions: The paper aims to provide a comprehensive and updated state-of-the-art review of 5G network slicing using SDN and NFV.Its scope includes service and business requirements, slicing concepts and use cases, architectures, initiatives, standardization, orchestration, and future directions.

2. 5G service quality and business requirements

5G service and business requirements span high data rates, ultra-low latency, availability, security, mobility, QoE personalization, energy efficiency, and support for diverse applications and industries. These requirements motivate adaptable and autonomous network management across heterogeneous infrastructures.

  • Applications and verticals: 5G networks must support diverse applications whose performance requirements differ across M2M, health, education, media, vehicles, and other verticals.Examples include telemedicine, immersive media, cooperative vehicles, and collaborative gaming.
  • Mobility: Mobility management must evolve beyond centralized and hierarchical protocols to provide mobility on demand for device- and service-specific requirements.The paper identifies enhanced handover procedures and topology-aware gateway selection and relocation as examples.
  • Security: 5G security must address new business, trust, and service-delivery models, increased privacy concerns, and threats to critical mission applications.The paper specifically mentions defending against denial-of-service attacks affecting smart grids, public safety, water distribution, and natural-gas networks.
  • QoE and business requirements: QoE differentiation requires personalized management because application types and media services have different KPI and quality requirements.The paper connects this differentiation with quality-based charging and pricing policies.
  • Energy efficiency: Energy efficiency requires dynamic resource allocation and adjustment of active network elements as traffic and user density vary.The paper notes that 4G base stations contribute 60%–80% of cellular network energy consumption and discusses small cells and relays as possible approaches.

3. 5G network softwarization and slicing: Concepts & use cases

5G network softwarization uses programmable and virtualized technologies to support flexible, isolated, customizable network slices across network segments and administrative domains.

  • 5G network softwarization: Network softwarization applies software programming to design, deploy, manage, and maintain network components and services across 5G infrastructure.It targets agility, cost-effectiveness, programmability, flexibility, adaptability, and end-to-end service management.
  • 5G network softwarization: 5G network segments—including RAN, transport, core, mobile edge, and network clouds—have distinct technical requirements and softwarization levels.The paper illustrates these technologies across the 5G architecture.
  • Mobile edge network softwarization: Mobile edge computing moves content, network functions, and resources closer to users through virtualized platforms based on SDN, NFV, and information-centric networking.MEC provides high bandwidth, low latency, location awareness, and real-time radio-network insight, while caching can reduce core-network processing traffic.
  • Core-network softwarization: 5G core and service-plane functions are expected to run as VNFs in virtual machines over standard Fog or cloud-computing servers.Slices can deploy network and service VNFs at different sites according to storage and latency requirements.
  • Transport-network softwarization: Programmable transport networks use SDN/NFV interfaces to accommodate services and implement resource discovery and optimization in the 5G control plane.This design adapts transport networks to future programmable RAN requirements.
  • Network slicing concepts: Network slicing provides E2E logical networks on shared physical or virtual infrastructure, with isolation, independent management, and on-demand creation.Slices may span access, transport, core, and edge domains, including components across different administrations.
  • Network slicing concepts: The NGMN slicing model separates service instances, network slice instances, and sub-network instances to represent services and their required network characteristics.A network slice instance may support one or multiple service instances and comprise shared sub-network instances.
  • Network slicing principles: Network slicing principles include automation, reliability, scalability, isolation, programmability, hierarchical abstraction, customization, and elastic resource use.These principles support management and orchestration across multiple segments, domains, and tenants.

4. 5G network slicing enabling technologies

SDN enables programmable, centrally controlled 5G networks and supports network slicing through controller-based orchestration and increasingly stateful data-plane technologies. However, OpenFlow-based traffic management can create scalability, reliability, security, and processing-delay concerns.

  • Software defined networking (SDN): SDN separates control and forwarding planes to provide centralized, programmable, network-wide orchestration for 5G services.Standardized southbound interfaces include OpFlex, FoRCES, and OpenFlow.
  • Software defined networking (SDN): The ONF slicing architecture represents each SDN client context as a potential network slice managed by an SDN controller.The controller applies rules or policies to manage slices.
  • Traffic management applications for stateful SDN data plane: OpenFlow provides a platform-agnostic interface between SDN data and control planes but supports only L2/L3 transport and lacks data-plane state maintenance.Packet states are maintained heavily by the external SDN controller.
  • Traffic management applications for stateful SDN data plane: Advanced data-plane technologies such as OpenState, P4, POF, SDPA, and SNAP expose persistent state and improve forwarding programmability.These technologies address limitations associated with static OpenFlow forwarding abstractions.
  • Traffic management applications for stateful SDN data plane: Advanced data-plane programmability supports resource slicing, isolation, and automated deployment of 5G services over programmable SDN infrastructure.It allows developers to exploit SDN data-plane resources for 5G applications.

4.3. Network function virtualization (NFV)

NFV virtualizes network functions on commodity hardware, decoupling software from hardware to enable flexible deployment and dynamic service provisioning. ETSI’s NFV architecture integrates infrastructure and tenant SDN controllers that provide distinct underlay and overlay abstractions.

  • Network function virtualization (NFV): NFV virtualizes functions such as firewalls, TCP optimizers, NAT64, VPNs, and DPI on commodity hardware.Network functions can be deployed, dynamically allocated, and associated with virtual network functions.
  • Network function virtualization (NFV): Service Function Chaining defines an ordered list of abstract service functions applied to classified packets, frames, or flows.
  • Network function virtualization (NFV): NFV decouples software and hardware, increases deployment flexibility, and enables dynamic operation and service provisioning.Operators can introduce tailored services and scale performance according to customer demands.
  • Network function virtualization (NFV): The paper notes that software-based SDN/NFV implementations must still be assessed against the technical performance requirements of different verticals.
  • NFV reference architecture: ETSI’s NFV MANO architecture includes infrastructure and tenant SDN controllers that centralize connectivity-related control at different levels.The architecture integrates SDN controllers into the NFV reference architecture.
  • NFV reference architecture: The Infrastructure SDN Controller manages underlying networking resources and can change NFV infrastructure behavior according to VIM specifications adapted from tenant requests.
  • NFV reference architecture: The Tenant SDN Controller dynamically manages tenant VNFs and forwarding-plane resources, operating in the tenant domain.It may be instantiated as a VNF or as part of the network management system and uses interfaces such as OpenFlow, NETCONF, and I2RS.
  • NFV reference architecture: The TSDNC provides a VNF-based overlay abstraction, whereas the ISDNC provides an underlay for VNF deployment and connectivity.The ISDNC is unaware of the slices and tenants using the VNFs it connects.

4.4. Multi-access edge computing (MEC)

MEC places cloud-computing capabilities and IT services at the mobile-network edge, enabling low-latency processing and local service delivery for 5G slicing. Its architecture combines MEC application servers, applications, services, and traffic offload functions.

  • Multi-access edge computing (MEC): MEC processes data near where it is generated and consumed, supporting ultra-low latency and interactive user experiences.It targets business-critical applications and busy venues such as shopping malls and train stations.
  • Multi-access edge computing (MEC): In 5G network slicing, MEC can enhance video-streaming QoS/QoE, optimize mobile resources, and turn access nodes into intelligent service hubs.Context-aware services can use RAN information such as location, cell load, and allocated bandwidth.
  • MEC architecture: The MEC application server runs on MEC NFVI and provides end-user services through individual MEC applications.MEC services are hosted on the MEC platform, which intermediates between applications and the platform.
  • MEC architecture: MEC service nodes may operate locally in a data center or remotely in the cloud, while applications and services interface with the Traffic Offload Function.The TOF prioritizes traffic through policy-based packet monitoring and redirection.

4.5. Cloud/fog computing

Cloud computing provides on-demand applications, platforms, and heterogeneous infrastructure, while network hypervisors abstract physical resources into isolated virtual slices. These abstractions support service delivery across diverse SDN providers and higher-layer network services.

  • Cloud computing: Cloud computing offers on-demand provisioning of applications, platforms, and heterogeneous infrastructure including servers, networks, and storage.The service-provider role is divided between infrastructure providers and service providers.
  • Cloud computing: SaaS provides hosted applications, PaaS provides an application development and management platform, and IaaS provides computing infrastructure as a service.
  • Network hypervisors: Network hypervisors abstract physical links, network elements, and control functions into logically isolated virtual network slices.They expose high-level abstractions and APIs for creating complex network services.
  • Network hypervisors: Network hypervisors can interconnect multiple SDN providers through a single abstraction, enabling applications to establish end-to-end flows without handling provider differences.
  • Network hypervisors: Through hypervisors, higher-layer services such as load balancing, firewalls, and L2/L3 link or network protocol services can be implemented.

4.7. Virtual Machines

This section contrasts VM-based and container-based virtualization while situating SDN and NFV within network softwarization for 5G slicing.

  • VMs virtualize physical resources, share computing, storage, memory, and network resources, and isolate guest operating systems from hosts and other VMs.
  • Containers provide lighter-weight OS-level virtualization, avoiding hardware indirection and enabling higher application density than VMs.
  • Highly mobile users may favor VMs because they can provide full logical isolation for VNFs operating within a network slice.
  • The section frames SDN and NFV as complementary foundations for flexible, automated, software-based 5G networking and slicing.
  • SDN decouples control and forwarding planes, whereas NFV runs network functions on commodity servers and can operate over legacy networks.

5. State-of-the-Art: 5G network slicing architectures and implementations

The paper surveys architectures and projects advancing SDN/NFV-based 5G slicing, emphasizing multi-service, multi-tenant, programmable, and multi-domain deployments.

  • The surveyed standard bodies, associations, and alliances coordinate 5G development and standardization around networks targeted for 2020 and beyond.
  • 5GEx develops an SDN/NFV-based multi-domain, multiservice orchestration platform combining end-to-end network and service elements across multi-vendor environments.
  • 5GEx: 5GEx uses slices as core infrastructure elements and supports collaborative models, higher-level abstractions, and third-party orchestration across exchange points.
  • MATILDA: MATILDA targets end-to-end orchestration of 5G-ready applications and services over sliced programmable infrastructure using unified slice creation and maintenance strategies.
  • 5G NORMA: 5G NORMA proposes multi-service and multi-tenant architecture based on adaptive network-function decomposition, allocation, and software-defined mobile control.
  • 5G NORMA: 5G NORMA functional blocks connect resource management, service requirements, mobility information, mobility-driven orchestration, and QoE/QoS reporting.
  • Across projects, multi-tenancy spans RAN, access, transport, computing, storage, switching, and transmission resource management across domains.

6.1. Open source orchestrators for network slicing

This section reviews open-source and industry orchestration platforms that automate, manage, test, and deploy virtualized 5G network resources, services, and slices.

  • Orchestrators automate creation, monitoring, and deployment of resources and services in softwarized and virtualized environments; ETSI distinguishes resource and service orchestration roles.
  • OSM: OSM follows ETSI NFV information models and combines service, resource, and configuration management for commercial NFV networks and network slices.
  • OpenMANO provides an ETSI NFV MANO realization focused on performance and portability through Enhanced Platform Awareness principles.
  • OpenNFV supports NFV and SDN infrastructure across open-source ecosystems, with automatic VNF deployment and monitoring through its NFV director.
  • CloudNFV combines cloud computing and SDN for multi-vendor NFV, with orchestration of VNF placement and service connectivity.
  • OPNFV accelerates multi-vendor NFV development through performance and use-case testing against current standard specifications.
  • M-CORD: M-CORD disaggregates and virtualizes RAN and core functions, while its ONOS-based slice manager monitors and dynamically scales operator services.

6.2. Global standardization efforts on 5G network slicing

Global standardization work addresses network-slicing requirements, architecture, management, orchestration, security, mobility, and end-to-end service support across network domains.

  • Industry discussions cover network-slicing concepts and requirements and assess impacts across network layers, including the core network and RAN.
  • ETSI: ETSI initiatives target full automation of 5G service deployment, configuration, assurance, delivery, and optimization, including end-to-end slicing management.
  • 3GPP: 3GPP assigns use cases and requirements to SA1, system architecture to SA2, security capabilities to SA3, slice management to SA5, and RAN slicing awareness to RAN1/2/3.
  • ITU-T: ITU-T IMT2020 supports dedicated logical networks with capability exposure, SDN/NFV-based softwarization, diverse QoS, and edge-cloud support.
  • ONF: ONF applies SDN architecture to 5G slicing and specifies TAPI for integrating transport-network control and monitoring with higher-level applications.
  • BBF: BBF coordinates bearer-network slicing requirements and interfaces with 3GPP while defining fixed-access network-sharing specifications.
  • IETF: IETF work includes general slicing requirements, architecture, lifecycle management, domain orchestration, and ACTN applicability to network slicing.

6.3. Proof of concepts (PoC) for 5G network slicing

Proofs of concept demonstrate 5G network slicing across RAN, multi-tenant, edge, cellular, C-RAN, and multi-domain settings, including dynamic resource allocation and deployment.

  • PoC coverage: PoCs cover RAN slicing, multi-tenant hybrid slicing, and 5G edge resource slicing.5G-EmPOWER demonstrates multi-tenancy support in a WLAN by assigning access-point resources per tenant according to traffic requirements.
  • Cellular and transport slicing: OVNES demonstrates the feasibility and reliability of overbooking network slices in real cellular deployments.
  • Cellular and transport slicing: An SDN-based C-RAN prototype shares spectrum bandwidth among slices while considering their requirements.
  • Multi-domain slicing: A multi-domain PoC illustrates on-demand creation and dynamic deployment of network slices across multiple domains.

6.4. Summary and lesson learned

Open-source orchestrators, standardization efforts, and PoCs collectively support the development and adoption of 5G network slicing across NFV and SDN ecosystems.

  • Open-source orchestrators, standardization efforts, and PoCs summarize industrial and academic progress in 5G network slicing.
  • Network slice orchestrators primarily manage resources and support NFV components and SDN infrastructure across different levels of 5G ecosystems.
  • OPNFV targets multi-vendor NFV performance and interoperability, while CloudNFV, Cloudify, and T-NOVA automate NFaaS operations on virtualized 5G infrastructures.
  • The surveyed PoCs and standardization efforts indicate continued development of network slicing through industry and standards-body contributions.

7. 5G network slicing orchestration and management

The survey reviews orchestration and management of 5G network slices across single and multiple administrative domains, including edge, fog, cloud, and RAN environments. It emphasizes multi-domain coordination, tenant-aware resource management, and unresolved monitoring and service-interface challenges.

  • 7.2. Multi-domain orchestration and management: Multi-domain orchestration enables automated slice provisioning across multiple technologies and operators, addressing scalability problems associated with single-domain managers.
  • 7.2. Multi-domain orchestration and management: Multi-domain management maps service requests across operators and technologies while matching each service’s ELA requirements.
  • 7.2. Multi-domain orchestration and management: 5G-PAGODA combines RAN, transport, core, and application enablers in slices that can operate over infrastructure managed by multiple providers.
  • 7.2. Multi-domain orchestration and management: Its resource orchestration module maintains a global domain view to place VNFs and create forwarding graphs across resources such as multiple data centers.
  • 7.2. Multi-domain orchestration and management: Multi-domain service conductors map service requirements to administrative domains through service-conductor and cross-domain modules.
  • 7.3. Edge, cloud, and fog management: Edge, cloud, and fog approaches support VNF placement or migration, on-demand multi-domain slicing, MEC brokering, and tenant-aware SLA/ELA resource allocation.
  • 7.3. Edge, cloud, and fog management: Popularity-based replication can reduce service time in service-function-chain slicing, while the surveyed approaches identify further needs for QoS/QoE monitoring and resource-management algorithms.
  • 7.5. Summary and lessons learned: The section compares orchestration features and RAN slicing approaches across research projects and standards bodies, with a goal of global service experience without separate provider agreements.

8. Future challenges and research directions

The paper identifies unresolved challenges in realizing 5G network slicing, spanning orchestration, resource management, security, RAN virtualization, and mobility. It outlines research directions for making slicing effective across services, tenants, domains, and network conditions.

  • SDN, MEC, fog/cloud computing, and NFV offer flexibility and agility, but substantial challenges remain before 5G network slicing can be realized.
  • Network sharing and slicing in 5G: Dynamic resource sharing can improve utilization, but requires intelligent scheduling plus effective network-function placement and intra- and inter-slice management.
  • Future work must address slice isolation, mobility management, dynamic slice creation, security, and standardized interfaces for Network as a Service.Isolation policies and mechanisms must operate across virtualization layers while preserving QoS/QoE under congestion and cross-slice interference.
  • End-to-end slice orchestration and management: End-to-end orchestration must coordinate slices holistically so each meets service and ELA/SLA requirements while using underlying resources efficiently.Adaptive decisions should account for current slice states, predicted future system states, and user demands.
  • RAN slicing remains immature because containers and virtual machines do not by themselves virtualize or isolate radio resources such as spectrum and hardware.
  • Mobility management in 5G network slicing: Mobility management must support slice-specific mobility and latency requirements, including service-aware QoS/QoE control and seamless movement across SDN controllers.

9. Conclusion

The paper surveys how SDN and NFV can support adaptable, programmable, and cost-effective 5G networks serving diverse vertical requirements. It consolidates enabling technologies, architectures, orchestration, standardization, and future research challenges.

  • SDN and NFV are presented as technologies for overcoming 5G resource-management and orchestration challenges while supporting different vertical requirements.
  • Network slicing is positioned as central to addressing vertical industries’ requirements for real-time capabilities, latency, reliability, security, and guaranteed ELAs/SLAs.
  • The survey covers 5G requirements, softwarization and slicing concepts, technology enablers, industrial initiatives, architectures, deployment strategies, orchestrators, proof-of-concepts, and standardization.
  • It also reviews single-domain and multi-domain slice management and orchestration, including multi-tenant, edge, and fog-network scenarios, before presenting future challenges and research directions.
Loading 1912.02802v1…