Source-linked AI summary
NFV: State of the Art, Challenges and Implementation in Next Generation Mobile Networks (vEPC)
Hassan Hawilo, Abdallah Shami, Maysam Mirahmadi, Rasool Asal
TL;DR
Mobile connectivity growth is increasing pressure on service providers to expand costly core-network infrastructure. The paper presents an NFV framework and a vEPC entity-grouping criterion, with analysis reporting a 70 percent reduction in network control traffic.
Problem
Growing mobile connectivity demand requires additional core-network infrastructure despite increasing CAPEX and OPEX investments.
Method
The paper introduces an NFV framework and groups related vEPC entities on single physical or adjacent devices based on their interconnections and functionalities.
Results
70 percent reduction in network control traffic was reported for the proposed virtualized-network-function grouping criterion.
Takeaways & Limitations
The proposed grouping reduces control-signaling traffic while maintaining flat architecture and separation of control and data planes.
Abstract
from arXiv · showhide
As mobile network users look forward to the connectivity speeds of 5G networks, service providers are facing challenges in complying with connectivity demands without substantial financial investments. Network Function Virtualization (NFV) is introduced as a new methodology that offers a way out of this bottleneck. NFV is poised to change the core structure of telecommunications infrastructure to be more cost-efficient. In this paper, we introduce a Network Function Virtualization framework, and discuss the challenges and requirements of its use in mobile networks. In particular, an NFV framework in the virtual environment is proposed. Moreover, in order to reduce signaling traffic and achieve better performance, this paper proposes a criterion to bundle multiple functions of virtualized evolved packet-core in a single physical device or a group of adjacent devices. The analysis shows that the proposed grouping can reduce the network control traffic by 70 percent.
I. INTRODUCTION
NFV addresses rising telecommunications costs and connectivity demands by separating network functions from proprietary hardware and running them on virtualized infrastructure. The paper introduces an NFV framework and discusses its deployment in mobile networks.
- Motivation: Rising mobile connectivity demand is increasing infrastructure investment while returns on network investments remain limited.The growth includes smartphones, tablets, laptops, sensor networks, and machine-to-machine connectivity.
- Motivation: Proprietary hardware dependence, energy costs, specialized personnel, and infrastructure-management complexity increase expenditures and limit innovation.These pressures also increase time-to-market for telecommunications services.
- Paper scope: The paper provides a guideline for designing and implementing NFV in telecommunications core networks and reviews its framework, requirements, use cases, and vEPC grouping.The proposed grouping targets reduced signaling traffic in virtualized EPC deployments.
- NFV concept: NFV separates network functions from proprietary hardware and runs them as software applications on commercial off-the-shelf platforms.Functions can be instantiated in datacenters, network nodes, or end-user premises as required.
- Expected benefits: NFV offers openness, scalability, flexibility, shorter development cycles, improved operating performance, and reduced CAPEX and OPEX.These benefits motivate the telecommunications industry's interest in NFV.
III. NFV FRAMEWORK
The NFV framework combines physical infrastructure, virtualization, cloud services, and orchestration to host and manage virtual network functions. Its proposed placement centralizes key management entities at the hypervisor level.
- Virtualized platform components: An NFV deployment consists of physical servers, a hypervisor managing physical resources, and guest virtual machines running applications.The guest VM emulates the architecture and functionality of a physical platform.
- Virtualized platform components: Virtual machines can run on high-volume servers in datacenters, network nodes, and end-user facilities.Cloud services provide on-demand computing resources in these environments.
- NFVI: NFVI uses infrastructure and networking services, including IaaS and NaaS, to provide the infrastructure for network-function virtualization.The paper notes that no standard definition of NaaS exists and that it is often considered part of IaaS.
- Framework management: The NFV framework requires dynamic VNF initiation and orchestration while managing resource allocation, dependencies, availability, and other hosting attributes.The framework also manages the NFVI hosting environment on IT virtualization technologies.
- Proposed placement: The proposed placement groups the Virtual Resources Manager, VNF Manager, and Orchestrator at the hypervisor level.This creates a centralized controller while the same infrastructure can host VNFs and other IT applications.
V. RECEPTION BY INDUSTRY
Industry interest in NFV has led telecommunications equipment vendors and IT companies to investigate and support its realization. The paper also identifies performance and security challenges that can hinder implementation.
- Industry reception: Major telecommunications vendors have begun adopting and upgrading equipment to support NFV.The named vendors include Ericsson, Nokia, Alcatel-Lucent, and Huawei.
- Challenges: NFV faces challenges that may degrade performance and hinder implementation in telecommunications.The paper presents requirements, challenges, and proposed solutions in this context.
- Security: NFV security should approach the level of proprietary hosting environments and must address virtualization, computing, infrastructure, and application domains.The paper recommends protecting the hypervisor and securing data communication and VM migration.
- Security: Programmable orchestration and infrastructure-interaction APIs introduce higher security threats to VNFs.The paper lists security challenges and proposed solutions in its challenge table.
B. Computing Performance
VNF performance depends on both underlying hardware and software design, requiring techniques that exploit parallelism, memory isolation, direct I/O, and processor affinity. These optimizations can conflict with conventional automated resource allocation.
- Computing Performance: Underlying processor architecture, clock rate, cache size, memory bandwidth, and speed strongly affect VNF performance.VNF software design also plays a major role in performance.
- Computing Performance: Multithreading and distribution across multiple cores or hosts can improve execution of high-demand VNFs.The paper presents these techniques as approaches for achieving high VNF performance.
- Computing Performance: Independent memory structures are used to avoid operating-system deadlocks, while private memory should be erased before reallocation.The security implementation separates and protects computing-domain resources.
- Security requirements: Secure virtualization and networking require hypervisor isolation, authenticated access, encrypted data, and secured techniques such as TLS or IPSec.These measures are organized across virtualization, computing, and infrastructure domains.
- Implementation techniques: Accelerated vSwitches, SR-IOV-compliant NICs, processor affinity, and direct I/O access are additional performance techniques.Processor affinity is intended to exploit cache memories.
- Implementation constraints: Implementing high-performance VNF techniques may require a different approach from automated resource allocation in current IT server pools.The paper identifies this as an implementation constraint.
C. Interconnection of VNFs
Virtualized network functions can be interconnected through multiple virtual-network scenarios, trading off performance, flexibility, and isolation. Placement on the same physical server or LAN determines the connection path.
- VNFs on the same physical server and LAN connect through the same virtual switch.
- VNFs on the same server but different LANs use a different virtual connectivity scenario.
- SR-IOV-compliant NICs provide faster, higher-throughput virtual-machine access than hypervisor-managed virtual interfaces.Hypervisor-managed interfaces are simpler to configure, while SR-IOV offers higher performance.
D. Portability
The virtual-environment approach improves VNF portability and resource isolation, while orchestration and standardized templates support automated management across the NFV infrastructure.
- Bare-metal VNF deployment offers predictable performance but sacrifices resource isolation and complicates software-instance security.Multiple software appliances run as processes on the same operating system.
- Virtual-environment deployment improves portability and presents hardware resources uniformly to VNFs.Each VNF can run on its specific operating system without awareness of the underlying operating system.
- Independent virtual machines managed by a hypervisor isolate VNFs and prevent unexpected interactions between them.Strict resource mapping is required to guarantee isolation.
- Standard templates for VNFs and computing infrastructure enable drag-and-drop deployment and automated management.
- The orchestration management system uses north-bound and south-bound interactions to manage VNFs, NFVI, policies, images, descriptions, and forwarding graphs.
F. Co-existence with Legacy Networks
NFV must coexist with legacy equipment and support carrier-grade reliability while addressing rising mobile-network demand through virtualization of base stations and core functions.
- VNFs should coexist with legacy equipment without disrupting management systems, forwarding graphs, services, or performance.The transition between VNF instances and physical functions should be secured and interruption-free.
- Carrier-grade NFV requires failure resilience, service continuity, and service assurance.Automated VNF reconstitution and near-real-time resource scaling support these requirements.
- NFV service models include NFVIaaS, VNFaaS, and VNPaaS, with use cases including fixed-access, content-delivery, and home-environment virtualization.
- Growing mobile-device and application demand requires operators to upgrade base stations and network cores for throughput, latency, and quality of service.
- Mobile base-station virtualization is initially considered in higher network-stack layers because physical-layer signal processing makes virtualization challenging.For LTE eNodeB, implementation is described first for layer 3 and then layer 2.
- Sharing remote base-station infrastructure can improve area coverage while minimizing CAPEX and OPEX investment.VNFs can also be upgraded to support multiple telecommunications technologies and new releases.
VIII. VIRTUALIZATION OF THE EVOLVED PACKET CORE (EPC)
The paper frames EPC as the target for mobile-core virtualization and proposes grouping its entities according to interactions and workload. The grouping divides entities into four segments and aims to reduce signaling and data-plane congestion while preserving EPC architectural principles.
- EPC is a simplified all-IP core architecture introduced in Release 8 to support mobile broadband and multiple access technologies.
- The LTE EPC includes MME, SGW, PGW, HSS, and PCRF entities supporting control, packet forwarding, external connectivity, subscriber data, and policy functions.
- vEPC entities are grouped according to their interactions and workload, with each group instantiated on one physical server or local network when beneficial.
- The grouping approach targets lower control-signaling traffic and data-plane congestion while maintaining flat architecture and control/data-plane decoupling.
- The proposed grouping divides EPC entities into four segments.The segments are summarized in Table 2 and illustrated in Figure 5.
A. Segment One
The proposed vEPC grouping colocates entities whose interactions benefit from internal processing while preserving EPC architectural principles. It combines control- or data-related functions to reduce network transactions and improve virtualized processing.
- Segment One: MME is grouped with the HSS front-end so authentication and authorization occur internally after one UDR query.The HSS FE temporarily caches user information and performs authentication and authorization with the MME.
- Segment One: LDAP carries HSS FE–UDR communication instead of Diameter, providing faster database querying with fewer resources.The passage describes LDAP as more efficient and faster than Diameter for database information queries.
- Segment One: SGSN is paired with HLR FE, but remains separate from EPC entities because its combined control and data plane conflicts with EPC decoupling.The grouping assumes continued support for 2G and 3G networks.
- Segment One: PGW is merged with SGW to reduce data-plane processing nodes and eliminate signaling transactions between them.The merge also supports centralized processing, CPU-cache efficiency, reduced Vswitch routing, and direct network-interface access.
D. Segment Four
The fourth segment groups policy, charging, user-data, and OSS/BSS-related functions, while the broader design uses GTP interfaces for small control-signaling exchanges. The quantitative analysis evaluates signaling traffic before and after these groupings.
- Segment Four: UDR, PCRF, OCS, and OFCS are migrated together to streamline policy generation, charging enforcement, and OSS/BSS interactions.UDR–PCRF colocation reduces information exchange and policy-generation latency, while grouping OSS/BSS-related entities limits interaction fragmentation.
- Segment Four: GTP is used between segments instead of Diameter because its UDP transport is suited to small control-signaling packets.The passage contrasts this with SCTP and TCP transport, whose connection-control packets can reduce performance for small exchanges.
- Segment Four: Most internalized transactions require much computational power despite reducing control signaling and network congestion.This is stated as a computational trade-off of the proposed grouping.
- Quantitative Analysis: 6.2 to 1 transaction(s) per subscriber is the reported decrease for MME–HSS signaling after grouping MME with HSS FE.The total transaction traffic in the cited scenario decreased to 173,239 transactions per second.
XI. CONCLUSIONS
The paper positions NFV as a response to telecommunications’ hardware dependence and concludes that its grouping criterion can reduce physical-network control traffic. The authors also identify technical challenges that may hinder NFV adoption.
- Conclusions: NFV decouples network functions from proprietary hardware and applies IT virtualization to telecommunications infrastructure.The conclusion frames this as NFV’s central industry objective.
- Conclusions: Technical challenges may hinder NFV’s integration into the telecommunications industry.The paper calls for IT organizations, enterprises, vendors, and researchers to explore approaches addressing these challenges.
- Conclusions: 70 percent reduction in network control traffic is reported for the proposed virtualized-network-function grouping.The grouping criterion is intended to minimize transactions on the physical network.