(China Telecom Beijing Research Institute,Beijing 100035,China)

Network Function Virtualization Technologyy:: Progress and Standardizatiioonn

Huiling Zhao,Yunpeng Xie,and Fan Shi

(China Telecom Beijing Research Institute,Beijing 100035,China)

Network innovation and business transformation are both necessary for telecom operators to adapt to new situations,but operators face challenges in terms of network bearer complexity,business centralization,and IT/CT integration.Network function virtualiza⁃tion(NFV)may inspire new development ideas,but many doubts still exist within industry,especially about how to introduce NFV into an operator’s network.This article describes the latest progress in NFV standardization,NFV requirements and hot technology issues,and typical NFV applications in an operator networks.

network functions virtualization(NFV);overlay network;virtual extensible LAN(VXLAN);service chaining

1 NFV Standardization Progress

1.1 ETSI NFV Progress

In October 2012,AT&T,British Telecom,Deutsche Telekom,Orange,Telecom Italia,Telefonica,and Ve⁃rizon established the Network Functions Virtualiza⁃tion Industry Specification Group(NFV ISG)in the ETSI.This group will define the specifications for architecture that supports NFV hardware and software and will create a guide to virtualized network functions.NFV ISG will cooperate with other standards organizations to consolidate existing virtu⁃alization technologies and standards.

NFV ISG intends to leverage standard IT virtualization tech⁃nology and consolidate many different types of network equip⁃ment into industry⁃standard,high⁃volume servers,switches and storage.Software with particular functions could be in⁃stalled or uninstalled on hardware in various locations in a net⁃work,and new equipment would not need to be installed.Bene⁃fits of NFV for network operators and customers include[1]:

·reduced equipment cost and power consumption

·lower capex and opex

·increased speed of deployment and provisioning of new net⁃work services

·increased investment margins for new services

·a virtual appliance market that is open to pure software en⁃trants

·encourages more innovation and new services for much low⁃er risk.

NFV ISG now has 184 members,including operators,net⁃work equipment vendors,IT equipment vendors,and technolo⁃gy vendors.The NFV ISG has a technical steering committee that manages four working groups and two expert groups.Dif⁃ferent working groups and expert groups focus on:

·architecture for the virtualization infrastructure,including infrastructure requirements in the computing,storage,and network domains

·management and orchestration,including NFV platform management functions such as network mapping for end⁃to⁃end services,allocation and expansion of hardware resourc⁃es,and VNF instance tracing

·software architecture,including the implementation environ⁃ment for VNF

·reliability and availability,including resilience and fault tol⁃erance through VNF load⁃allocation approaches and VNF in⁃stance portability

·security of NFV platforms

·performance and portability,including scalability,efficien⁃cy,and migration performance,from dedicated platforms to general⁃purpose hardware.

In 2013,NFV ISG focused on designing high⁃level docu⁃ments.It has released NFV use cases,requirements,architec⁃ture,terminology,proof of concept(PoC),and other technical documents as well as NFV White Paper V1.0 and NFV White Paper V2.0.The focus of NFV ISG has shifted from identifying requirements to defining them,and NFV ISG is attempting to achieve feasible results by specific deadlines.In the first half of 2014,NFV ISG has focused on PoC and is looking to collect

and evaluate products and prototypes that satisfy NFV require⁃ments.This will help promote NFV development.By the end of February 2014,nine PoC proposals had been accepted.

Not long ago,NFV ISG also released NFV Phase 2 discus⁃sion draft,which specifies the work plan for the first two years of NFV ISG.In this draft,two points should be noted.First, NFV ISG plans to establish an NFV steering board(NSB), which will be a major organizational entity focused on promot⁃ing NFV work.Compared with NFV ISG TSC,the NSB will have a more rights.The NSB not only coordinates technologies, as the current TSC does,but it also supervises the progress of NFV ISG.Second,an ad hoc group will be established to re⁃place the existing working groups and expert groups.Accord⁃ing to the NFV Phase 2 plan,the objectives and tasks of this ad hoc group will be specified by the fourth quarter of 2014, and the ad hoc group will begin work in 2015.

1.2 Network and NFV Standardization:CCSA Efforts

The China Communications Standards Association(CCSA) pays much attention to network and NFV standardization and guides NFV study and application in China.Software and virtu⁃alization have become important trends in the evolution and de⁃velopment of future networks.These two topics are highly com⁃plementary,and relevant representative technologies and proto⁃cols will be the basis of future networks.However,software⁃de⁃fined networking(SDN),NFV technologies,and the architec⁃ture of future networks are all still being studied and depend on the development of relevant technical standards.

TC1 of the CCSA focuses studies standards related to IP and multimedia communication.A large amount of study is being done on the virtualization of data centers,CDNs,and broad⁃band bearer networks.These standards have been completed:

·scenarios and requirements of future data networks(FDNs) (industry standard)

·general requirements of internet data center based on virtu⁃alization technologies(industry standard)

· router virtualization technical requirements(association standard)

·impacts of network edge virtualization on MAN(study sub⁃ject).

Other industry standardization projects that have been initi⁃ated include:

·scenarios and requirements of stream⁃specific state migra⁃tion in cloud data centers

·orchestration scenarios and technical requirements of FDN services based on cloud computing management platforms

·application scenarios and technical requirements of FDN⁃based CDN

·application scenarios and technical requirements of FDN⁃based broadband customer networks

·technical requirements of FDN⁃based broadband network ac⁃cess servers.

TC3 of the CCSA has established the Software Virtualization Network(SVN)Work Group and has been studying NGN key technologies,equipment,signaling protocols,and network ar⁃chitecture evolution.TC3 is a major technical work committee for telecommunications network architecture.The design of fu⁃ture network architecture and relevant technologies is within the scope of this work group.Intelligent communications net⁃work technology,which is a particular focus of TC3,is the foundation of future networks and will profoundly affect the fu⁃ture development the entire ICT network.

The CCSA TC3 SVN group,also called the Software Intelli⁃gent Communications Network Work Group,studies the archi⁃tecture and key technologies of future SDN and NFV networks. CCSCA TC3 SVN undertakes relevant standardization and pro⁃vides important references for development in this field.

One of the main subjects of TC3 SVN is the requirements, frameworks,and key technologies of SDN⁃based intelligent communication networks.This subject encompasses:

·general requirements of SDN⁃based intelligent communica⁃tion networks

·perception analysis in SDN⁃based intelligent communica⁃tions networks

·traffic scheduling in SDN⁃based intelligent communications networks

·policy control in SDN⁃based intelligent communications net⁃works

·evolution of existing networks to SDNs.

The other main subject of TC3 SVN is the requirements, frameworks,and key technologies of network virtualization. This subject encompasses:

·general requirements of network virtualization

·virtualized network functions

·virtualized network services

·virtualized evolution of existing networks.

TC3 SVN has also studied the requirements,frameworks, and key technologies of future networks.It has initiated study on the general technical requirements of SDN⁃based intelligent communication networks,technical requirements of SDN⁃based intelligent perception systems,control⁃plane platform virtualization of core networks,and technical requirements of SDN/NFV⁃based virtualized IMS.

2 Hot Technology Issues in NFV

NFV partly borrows from existing network virtualization technologies and also incorporates new technologies,such as software virtualization and SDN.NFV properly abstracts, splits,and schedules network function sets and involves many technologies.In this paper,only overlay network,virtualized traffic scheduling,virtual cluster,and networking technologies are discussed.

2.1 Overlay Network Technologies

Overlay network technologies are used to implement virtual⁃

ization over existing network architecture,and the basic over⁃lay network is not greatly changed.Thus,application bearers can be established in the overlay network and are separate from other network bearers.At present,overlay network tech⁃nologies are mainly used for high⁃volume interconnection in the internal networks of datacenters.Here,we describe main⁃stream overlay network technologies.

2.1.1 Virtual Extensible LAN(VXLAN)

VXLAN[2]is an important virtualization technology and subset of IETF standard drafts.VXLAN enables network virtu⁃alization by using MAC⁃in⁃UDP encapsulation to overlay a lay⁃er⁃2 network onto a layer⁃3 network.Each VXLAN is identi⁃fied with a 24⁃bit VNI.VXLAN encapsulation enables the layer⁃2 to communicate with any end point as long as the end points are in the same VXLAN segment.These end points may not necessarily be in the same IP subnet,so the problem of limited MAC address capacity in switches is eliminated.

2.1.2 Network Virtualization Using Generic Routing Encapsulation(NVGRE)

NVGRE uses the GRE tunneling protocol encapsulation,de⁃fined in RFC 2784[3]and RFC 2890[4],to create an indepen⁃dent virtual layer⁃2 network.In NVGRE,address learning is implemented by the control plane,but NVGRE has previously had no specific implementation solution for address learning until now.Compared with VXLAN,NVGRE is defective in terms of load sharing,i.e.,NVGRE cannot implement GRE key⁃based load sharing.In addition,NVGRE tunnels are end⁃to⁃end,so the number of tunnels increases exponentially as the number of terminals increases.As a result,the overhead for tunnel maintenance becomes very large.

2.1.3 Stateless Transport Tunneling(STT)

STT is also an overlay technology used to create a layer⁃2 virtual network over a layer⁃2 or layer⁃3 physical network[5]. In technical terms,STT is very similar to VXLAN.Tunnel end points of STT are also provided by a hypervisor vSwitch; VNIDs of STT are also 24⁃bit;and STT has a multipath advan⁃tage by controlling transmission source packet headers.The difference between STT and VXLAN is that STT fragments da⁃ta frames before encapsulation.Thus,the hardware accelera⁃tion of network cards can be fully utilized for higher efficiency. In addition,STT disguises STT packets as TCP/IP packets, and TCP packet headers do not maintain TCP state informa⁃tion;thus,re⁃transmission does not occur after packet loss.In this way,STT tunnels are less reliable.

2.2 Virtualized Resource Scheduling Technologies

Virtualized resource scheduling technologies use SDN and NFV to virtualize and intelligently schedule network traffic, service functions,and other resources.Such technologies are mainly used for virtualized traffic scheduling and service chaining.

2.2.1 Virtualized Traffic Scheduling

Virtualized traffic scheduling overcomes the limitation of a distributed IP network routing by using virtualization technolo⁃gies.It uses centralized route computing and traffic scheduling to dynamically balance traffic and optimize the architecture across the whole network.Virtualized traffic scheduling is mainly used in IP backbone networks to determine 1)how to define the abstraction of IP route function sets,2)the imple⁃mentation mode of centralized route decision systems,3)the re⁃liability of centralized systems,and 4)the real⁃time algorithms used to compute protection paths in this mode.The current trend for virtualized traffic scheduling is SDN and other new technologies,e.g.,adding a PCE/controller system to imple⁃ment a centralized route⁃decision system.

2.2.2 Service Chaining

Virtual firewalls,load balancers,gateways,and other service⁃processing functions in a network are called service function points.By processing traffic at a series of service⁃function points,a service chain is formed.This process is called service chaining[6].Unlike virtualized traffic scheduling,service chaining focuses on server programming for controlling traffic forwarding in a virtual network.Because it has been promoted by SDN and NFV,service chaining has received much atten⁃tion,and it is widely considered to have good prospects.

2.3 Virtual Cluster Technologies

A virtual cluster is formed when virtualization technologies are used to logically combine network elements(NEs)or their internal components in order to meet operational and manage⁃ment requirements.Currently,the study of virtual cluster tech⁃nologies is focused on homogeneous and heterogeneous virtual clusters.

2.3.1 Homogeneous Virtual Clusters

By expanding the control plane,a homogeneous virtual clus⁃ter virtualizes multiple physical devices of the same type into a single logical device.The cluster implements resource sharing and flexible scheduling in these physical devices through a re⁃source controller.By means of pooling,the virtual cluster has a uniform control plane and management plane and uses a unique ID.Compared with the original physical devices,the virtual cluster has much greater capacity and is much more re⁃liable.This technology is mainly used in backbone networks to solve the problem of insufficient single⁃server forwarding and insufficient throughput in core nodes.Moreover,it can be used for multi-service edge(MSE)pooling in IP networks and mo⁃bility management entity(MME)pooling in core networks.

2.3.2 Heterogeneous Virtual clusters

A heterogeneous virtual cluster consolidates different types

of physical devices in distributed mode.Thus,the number of managed or configured NEs and NE types is reduced,and ser⁃vice and network deployment can be made more flexible and efficient.At present,research is focused on virtual clusters of access control devices and switches,optical network unit (OLT)access control devices and home gateways,and routers and optical transport network(OTN).

3 Typical Applications of NFV

In the face of market competition,operational require⁃ments,and increasing maintenance costs,network operators have begun to explore NFV and have attempted to it to satisfy specific service requirements in data centers,mobile core net⁃works,and home networks.

3.1 Data Center Network Virtualization

Data center network virtualization[7]comprehensively shields underlying physical network appliances in overlay mode.In a virtualized data center network,physical network resources are shared,and different tenants are isolated by soft⁃ware or programming.Each tenant has a separate network defi⁃nition,including networking,traffic control,and security man⁃agement.A cloud data center resource⁃management platform is connected to an SDN controller through API interfaces.By means of programming,a multitenancy network can be flexibly deployed,and inter⁃datacenter deployment is also possible. Fig.1 shows datacenter network virtualization[8]-[10].

Data center network virtualization does not depend on under⁃lying networks so that security,traffic,and performance poli⁃cies can be flexibly implemented for different tenants.The net⁃work can also be automatically configured because of the pro⁃gramming capabilities[11].After overlay network technologies are introduced,however,the network architecture becomes more complicated,and the physical network cannot perceive the logical network.In addition,network performance is com⁃promised because the logical network is controlled by software.

▲Figure 1.Data center network virtualization.

3.2 EPC NE Virtualization

Virtualization of evolved packet core(EPC)network in⁃volves the use of a three⁃layer application+controller+switch architecture.Control functions,including traffic flow and traf⁃fic processing,are implemented by applications+controller layers.The switch layer implements stream⁃based forwarding functions or even integrates DPI and other traffic analysis and processing functions.Control⁃plane NEs are gradually central⁃ized,and a virtual control cloud is formed in the mobile core network by converging the System Architecture Evolution (SAE)gateway signaling plane with the MME or policy and charging rules function(PCRF)[7].Fig.2 shows EPC NE virtu⁃alization[12].

EPC NE virtualization unifies the network hardware archi⁃tecture with the NFV technology,so that the cost will not in⁃crease greatly due to increasing capacity.By separating ser⁃vice control from forwarding and separating software from hard⁃ware,EPC NE virtualization allows flexible service deploy⁃ment and enhancement,and thus reduces CAPEX and OPEX for network operators.

3.3 Home Network Virtualization

Home network virtualization separates control plane func⁃tions and service processing functions(such as firewall,ad⁃dress management,device management,and fault diagnosis)of home gateways(HGs)and set top boxes(STBs)in home net⁃works,and migrate these functions to the controller side or cloud end after virtualization.On the HGs and STBs,only physical interfaces and data plane layer 2 forwarding functions are remained[7],[13].Fig.3 shows an application scenario of home network virtualization.

▲Figure 2.Core network single⁃NE virtualization.

▲Figure 3.A home network virtualization scenario.

Home network virtualization simplifies end⁃user premises. Network operators can provide remote network fault diagnosis without continuously maintaining and upgrading STBs and HGs.Thus,services are more manageable,and less power is consumed.Home network virtualization also makes service de⁃ployment more flexible.Operators can deploy new hardware or software quicker and easier so that the time to market is re⁃duced[14].

4 Conclusion

NFV has succeeded in the IT industry and has entered the operator landscape.NFV has many advantages in various sce⁃narios and is a growing trend in the telecom industry.

However,NFV is still being standardized;relevant technical standards are not yet complete and require further in⁃depth study.Take service chaining for example.The functional points and logical combination sequence of service chaining may differ for different services.Therefore,it is general pur⁃pose service chaining applications urgently need to be defined. The content of NVF will improve as ETSI and CCSA continue with their study and formulation of relevant standards.

[1]ETSI.(2013,Oct.17).Network Function Virtualization—Introductory White Pa⁃per[Online].Available:http://portal.etsi.org/nfv/nfv_white_paper.pdf

[2]NFV⁃INF Network Domain Interworking⁃Data Plane,ETSI NFV INF(13) 000056,Aug.2013.

[3]Generic Routing Encapsulation(GRE),IETF RFC 2784,Mar.2000.

[4]Key and Sequence Number Extensions to GRE,IETF RFC 2890,Sept.2000.

[5]Vishwas Manral.(2012,Mar.22).Stateless Transport Tunneling(STT):Yet anoth⁃er cloud encapsulation or next⁃generation VxLAN?[Online].Available:http:// h30507.www3.hp.com/t5/HP⁃Networking/Stateless⁃Transport⁃Tunneling⁃STT⁃Yet⁃another⁃cloud/ba⁃p/109559

[6]Network Functions Virtualisation(NFV)Use Cases,ETSI GS NFV 001 V1.1.1, Oct.2013.

[7]Scenarios and Requirements of Future Data Network,CCSA Industry Standard, 2013.

[8]Xuan Luo,Baoqing Huang,Jianwen Wei,and Yaohui Jin,“Data⁃Center⁃Orient⁃ed SDN,”China Education Network,no.100,pp.24-27,Aug.2013.

[9]Qian Wang,Huiling Zhao,and Yunpeng Xie,“Standardization and Deployment of SDN,”ZTE Technology Journal,vol.19,no.5,pp.2-5,Oct.2013.

[10]N.McKeown,T.Anderson,H.Balakrishnan,G.Parulkar,L.Peterson,J.Rex⁃ford,S.Shenker,and J.Turner,“OpenFlow:enabling innovation in campus networks,”ACM SIGCOMM Computer Communication Review,vol.38,no.2, pp.69-74,Apr.2008.doi:10.1145/1355734.1355746.

[11]Baohua Lei,Feng Wang,Qian Wang,Heyu Wang,Yunpeng Xie,and Fan Shi, Deciphering SDN:Core Techniques and Practical Guide,Beijing,China:Pub⁃lishing House of Electronics Industry,2013.

[12]SDNAP.(2013,Oct.20).Architecture of ETSI NFV[Online].Available:http:// www.sdnap.com/sdnap⁃post/2856.html

[13]High Level Requirements and Framework for SDN in Telecommunication Broad⁃band Networks,BBF SD⁃313,Mar.2013.

[14]Network Located Residential Gateway,BBF PD⁃295,Oct.2012

Manuscript received:2014⁃03⁃05

Biograpphhiieess

Huiling Zhao(zhaohl@ctbri.com.cn)is director of Cloud Computing Research Cen⁃ter and chief engineer of Beijing Research Institute of China Telecom Corporation Limited.She is also the executive director of China Institute of Communications, chairman of the Information and Communications Network Technology Professional Committee,vice president of Beijing branch of China Institute of Communications, chairman of the Networking and Switching technology Committee of CCSA,and a member of the MEF Board.She has previously been granted special government al⁃lowances.She is one of the experts leading broadband network projects and tri⁃net⁃work integration projects in China’s 12th Five Year Science Plan.

Yunpeng Xie(xieyp@ctbri.com.cn)is a senior engineer of the Network Architec⁃ture and Cutting⁃Edge Technology Study Group at the Network Technology Depart⁃ment of Beijing Research Institute of China Telecom Corporation Limited.His re⁃search interests include SDN/NFV and future networks.He has received one provin⁃cial award,and submitted four patents applications.He is a joint author of two monographs,and has published more than 10 papers.

Fan Shi(shifan@ctbri.com.cn)is director of the Network Architecture and Cutting⁃Edge Technology Study Group at the Network Technology Department of Beijing Re⁃search Institute of China Telecom Corporation Limited.He is also the co⁃chair of MEF China Working Group and the leader of CCSA TC3 SAV working group.His research interests include SDN/NFV and next⁃generation internet.