NFV Network Functions Virtualization Virtualization of network functions

Premises, concept and architectural features of NFV

Evolution of architecture of networks

Capacity and capacity of communication networks and data transmission grew, and possibilities of networks developed during rather long time. However, by 2019 it is more and more noticeable that increase in capacity of networks does not get on well behind more and more rapid growth of need of the market for bandwidth and at new services. According to Cisco Visiual Network Index, network traffic grows exponential with the annual average growth rate of 26% till 2022, and no signs of decrease in rates of this growth are predicted further.

Figure 1. Growth of network traffic of 2017-2022 (source: Cisco)

It forces telecom operators and providers of services the Internet to expand the networks more and more in high gear, meanwhile, profitability from services grows far not so quickly as the need for bandwidth. Operators try to reduce costs for development of network and also to increase the speed of development and deployment of new features for revenue growth.

Traditional methods of development of networks consist in the "hardware" principle: "the device – function". In most cases to enter the next function, it is necessary to install other device. However, such methods of development of networks and an output to the market of new services by means of new physical devices face a number of restrictions.

Let's consider briefly these restrictions.

• Restrictions of flexibility

The proprietary vendor design of the equipment is that each network device has the fixed combination of the equipment and the software with small variations. It very much limits operators in the choice of functional combinations and opportunities of the equipment which can be unrolled on network. However, all growing market demands to different services and services do use of resources of such vendor solutions inefficient.

• Restrictions of scalability

Scaling options of physical network devices are limited as regarding the equipment, and the software. The equipment requires a power supply and the place for placement that is often limited in zones of dense building. On the other hand, possibilities of the available physical devices often do not keep up with growth rate of need for modern network functions. Each device is designed with a certain limit of opportunities, and at its achievement, it is necessary to install the new device which will remain underloaded for a long time.

• Long time of an output of services for the market

Market demands grow, but growth of number of functions and items of equipment are not always capable to keep up with these requirements. Often input of new features requires the upgrade existing and input of the new equipment. For this purpose careful development of migration process, confirmation of correctness and practical tests of a new solution is required. Assessment of the new equipment, reorganization of network topology, and, perhaps, involvement of new equipment suppliers who have required devices and functions in the portfolio is required. All this considerably increases capital costs and cost of ownership of infrastructure of network, leads to restriction of business and loss of potential income if competitors appear quicker.

• Administration restrictions

Monitoring systems use standard protocols, such as SNMP (Simple Network Management Protocol), NetFlow, Syslog, etc. for collection of information about a status of devices. However for monitoring of vendorospetsifichny (proprietary) parameters, the standard systems can be insufficiently. In this case, each network domain constructed on the equipment of a certain supplier requires a proprietary monitoring system. It also makes the contribution to capital costs and cost of ownership, multiplying "zoo" of such devices at the operator, requires personnel training and contents in staff of specialists with certificates of the relevant vendor.

• Operating expenses

The Vendorospetsifichny equipment raises operating expenses as requires existence in staff of the operator of the corresponding specialists. Or, as the option, requires use of so-called 'managed services' (i.e. "professional services" of the supplier). It leads to "vendorozavisimost", i.e., to attachment to solutions of the specific supplier. On the other hand, use of the equipment of many suppliers on network causes compatibility issues of the diverse equipment and raises operating expenses even more.

• Migration

Devices and networks should be serviced and upgraded periodically. It requires physical access to network elements and departure of technical personnel on installation sites to deploy the new equipment, to reconfigure physical links, and to carry out works with the engineering equipment of the website. It creates cost a barrier to decisions on migration and upgrade of network, slowing down the offer of new services to users.

• Smoothness of growth of capacity of network

It is difficult to predict requirements to growth of capacity of network (both on short, and on long terms). As a result, networks often are under construction with a large supply on capacity in calculation not explosive growth to the needs for traffic. According to estimates of experts, more than 50% of capacity of network remain nedovostrebovanny бòльшую a part of time. It conducts to unproductive capital costs. On the contrary, when resources of network are exhausted, there passes considerable time before the capacity of network manages to be expanded. Thus, development of network happens intermittently, to alternation of the periods of insufficient and excess capacity of network.

• Interaction

Often happens so that for acceleration of an output of the equipment to the market suppliers equip it new to functions before full completion of process of their standardization. In many cases it results in incompatibility of the diverse equipment on network, need of tests of network solutions for laboratories of operators and service providers before unrolling these devices on network. Then, the period of completions by equipment supplier and re-testing usually follows. And even in this case a part of not compatibility does not manage to be revealed, and it "emerges" already during operation the equipment on "live" network.

Concept of virtualization of network functions

Virtualization – technology which allows to start several operating systems on one physical server. The concept of virtualization of servers is for a long time used in data centers (DPC, data processing centers). At the same time, physical servers are replaced with their virtual "copies" working over hypervisors. It allows to reach, except other, more effective use of physical resources of data center.

NFV expands the concept of virtualization out of limits of servers, on all types of network devices. NFV can be defined as a method and technology which gives the chance to replace physical network devices with the certain functions on program entities performing the same functions on the public server hardware.

NFV is used as the "umbrella" term for designation of an ecosystem which consists of virtual network devices, management tools and infrastructure which integrates program entities with a normal computer hardware.

Figure 2. Transition from traditional infrastructure of network to virtualized (a source: Rajendra Chayapathi, Network Functions Virtualization (NFV) with a Touch of SDN. © 2017 Pearson Education, Inc.; TAdviser, 2019).

Actually, NFV separates the software from the equipment, and gives an opportunity to use any commercially available, normal equipment of COTS (Commercial Off the Shelf) for accomplishment on it specialized network functions which can be changed quickly and at any time.

Structure (Framework) of NFV

The term NFV was for the first time entered by chief operators of communication of the world on the congress SDN OpenFlow World Congress in 2012, chief operators of communication of the world. They analyzed the restrictions of a traditional method of development of network stated above and created the working group on development of the NFV ISG specifications (Industry Specification Group) under the direction of the European institute on development of standards for telecommunications of ETSI (European Telecommunications Standards Institute).

The working group of ISG put forward three main criteria which should be implemented in standards (recommendations) for NFV:

• Department (Decoupling): complete separation of the equipment and software.
• Flexibility: the automated and scalable deployment of network functions
• Dynamic transactions (Dynamic operations): control of operational parameters of network through exact (granular) management and condition monitoring of network.

On the basis of these criteria the generalized architecture of NFV shown in the drawing below was developed.

Figure 3. The generalized architecture of NFV (the Source: ETSI, TAdviser).

The architecture of NFV consists of three main subsystems:

• The virtualized network NFV functions (Virtualized Network Function)
• Infrastructure of virtualization of NFVI (NFV Infrastructure)
• Management subsystem and orchestrations of MANO (Management and Orchestration)

Except NFVI, subsystems represent the software, but not the equipment.

NFVI includes as the physical equipment (calculations, storage, network), and virtual "equipment": servers, storage systems, network devices. Virtualization level (hypervisors and guest operating systems) give the chance to turn VM virtual machines (Virtual Machines) which perform any functions ordered them on physical servers. Is not of great importance on what physical server it is unrolled for VM and works. Moreover, VM can move (to migrate) from one physical server on another without interruption of their work.

Strictly speaking, it is also necessary to include in architecture of NFV a subsystem of support of transactions and business OSSBSS (/) which is a part of traditional network system of the telecom operator. However, existence of this subsystem in architecture of NFV is time as operators cannot refuse in one step the existing OSS/BSS and at once pass to MANO (it possibly only for new networks of the operators starting the business from scratch and beginning to build network).

MANO should have unobstructed sight (an operational status, statistics of use and so forth) all program entities unrolled in the NFV system and to manage them. Therefore MANO submits the most suitable interface for OSS/BSS subsystem regarding collecting of operational data. In the future, in process of transformation of network, all OSS/BSS functions should pass to MANO.

Interaction of elements of architecture of NFV

The generalized blocks of architecture of NFV, in fig. 3 consist, in turn, of functional modules of lower level. For example, the administrative MANO block is a combination of three main functional modules: manager of infrastructure of virtualization of VIM (Virtualized Infrastructure Manager), manager of the virtual VNFM network functions (Virtualized Network Function Manager) and orkestrator of NFVO (NFV Orchestrator).

In architecture reference points (reference points) between functional units and modules via which they interact among themselves are also defined. It is necessary to notice that reference points are not interfaces (program or hardware). These points represent specifications of that information which should be transmitted through them and also where and as it should be processed.

In the drawing below reference points are shown by greasy circles with trunk lines. Their more detailed description can be found in the description from ETSI, or here.

Figure 4. More detailed architecture of NFV according to ETSI

Service lives in NFVI are subdivided into three categories:

• Equipment of servers
• Equipment of storage systems
• Equipment of network interfaces

The computing equipment of servers includes central processor CPU and memory which can be distributed between computer nodes using technologies of a clustering.

Storage systems can be local NAS (Network Attached Storage) and also connected on SAN technology (Storage Area Network).

Network equipment consists of a set of cards of network interfaces and ports which can be used by the virtual VNF network functions.

Any of these types of the equipment is not specially created for accomplishment of these or those functions, and represents public COTS hardware devices. All types of the equipment (processors, memory, storage systems, network interface cards and so forth) are united in the general pool. Separate devices from this pool are used as required for creation of VNF, and, after function completion of work, are released again.

Figure 5. The principle of virtualization in NFVI.

Functional units can stretch between different computer nodes in data center and even between data centers. So they are usually not concentrated in one network point, location or a point of presence on POP network (point of presence), and distributed.

Such distribution – very important quality of architecture of NFV which gives many advantages.

The functional unit of NFVI provides hardware resources in the form of VM virtual machines (Virtual Machine) on which VNF work. Virtual machines are a virtual analog of the physical equipment which is provided for start on them of VNF in the same way as if it was the physical equipment.

The manager of infrastructure of virtualization of VIM (Virtualized Infrastructure Manager) which is a part of MANO is intended for management of NFVI. He is responsible for resource management of servers, storage systems and network and also the software of Level of virtualization. VIM has the complete information about existence and employment of resources (inventory) and also attributes of their work (for example, power supply, processor busy condition and so forth), and besides, monitors of operation parameters (for example, statistics of use).

The manager of VNF – VNFM is engaged in creation and management of work of virtual network functions. He, at the request of Orkestrator NFV on formation of virtual function of VNF, requests appropriate resources from VIM and starts on them this VNF, or makes composite function of other, more elementary VNF. In the course of its work if expansion (scaling) of power of VNF, VNFM is required requests additional resources from VIM, receives them and adds to the working VNF (for example, adds a CPU processor core). It can be start of additional VM virtual machines, accumulation of amount of memory or space of storage system, connection of new network interfaces. As VIM knows "accounting system" (inventory), it can define whether there are cash COTS resources for satisfaction of a request, or a system approached a limit of their use. In the latter case various measures which enter functionality of ensuring flexibility of a system can be undertaken.

Figure 6. An example of selection of the additional CPU resources for VNF1.

VNFM manages the virtual VNF network functions via virtual devices of management of EM (Element Management, see fig. 4). These devices are necessary to provide interaction of VNF as well with the physical network PNF functions (Physical Network Functions) of normal physical devices of network of the operator which are inherited from traditional network. For the purpose of simplification of PNF in drawings are not shown.

Scope of responsibility of EM is similar to a traditional management system and administrations of the EMS elements (Element Management System). In traditional, physical network of the operator it is intermediate between the management system of NMS network (Network Management System) and physical elements of network performing network functions. EM interacts with VNF under the specialized protocols which remained from traditional network, and with VNFM – by means of the open protocols determined by ETSI. In completely virtualized networks need for EM, most likely, will disappear since they will be integrated with VNF.

In evolving to NFV network which contains both parts – traditional and virtualized, EM are necessary as well for universal accomplishment of the FCAPS functions (Fault, Configuration, Performance, Accounting, Security), i.e. for management at failures (fault), configuring (configuration), measurements of operation parameters (performance), accounting (accounting) and security (security).

In completely virtualized network of the FCAPS function will enter entirely an area of responsibility of VNFM.

Upon transition from physical network elements to virtual, telecom operators usually are not eager to change or transform those management tools network to which they got used and which well work with the traditional OSS/BSS systems. The structure of ETSI considers it, and does not demand from operators who take a way of transformation of NFV, in one step to change all developed network management system. So the existing systems of board can continue to work even in that case when physical network elements are replaced with VNF. However, the existing OSS/BSS systems have the shortcomings which do not allow them to use fully all advantages of NFV and also they cannot communicate directly with control blocks of NFV – VNFM and VIM. Of course, the existing systems can be adapted somehow for use of advantages of NFV, such as efficiency of work and elasticity. But it is not optimal approach because such systems were, as a rule, projected by the principle of "propriyetarnost" which do not allow to manage open platforms.

For the solution of this problem in structure of MANO there is one more functional unit under the name Orkestrator NFV, NFVO. It allows traditional OSS/BSS to manage transactions in NFVI and VNF through VNFM and VIM (See fig. 4).

The role of NFVO is not so obvious as VIM and VFNM, and, at first sight, it looks just as the buffer between them and traditional OSS/BSS. However, the orkestrator of NFVO plays very important role in structure of NFV ETSI. It manages complex deployment of services on network, makes a global picture of virtualization of services and exchanges this information with VIM and VNFM for deployment of services. In particular, shown in fig. 6 "A request for expansion of capacity of the processor from VNF1", arrives from NFVO.

NFVO also works with VIM and has the broad picture of cash of resources which it manages. In one NFV system there can be many VNFM and VIM, but coordinates their work of NFVO.

Above only the simplified picture of system operation of NFV as the format of article does not allow to consider all aspects in more detail is described.

Advantages of NFV

At the beginning restrictions of traditional methods of development of communication networks were briefly described. Let's consider how virtualization of the network NFV functions solves the majority of these restrictions and also introduces additional benefits. Much of what in traditional network of the operator was impossible and therefore such possibilities were not even considered becomes possible in NFV.

• Freedom of choice of the equipment

As NFV uses normal, commercially available computer hardware of COTS, operators can select the most suitable at the prices and support the equipment from numerous producers, and thus, most optimum to build the networks, as on costs, and functionality.

It is possible to tell that all variety of the traditional network equipment supplied by vendors in NFV comes down only to three of its types: server, storage system, network devices. However, the number of suppliers of such normal equipment are much more, than specialized telekomovsky. All functionality at the same time is provided with program functions which work at this limited nomenclature of the equipment.

Process of modification of the traditional equipment on network of the telecom operator usually is very long and costly. With NFV, operators can enter new features for several clicks from a control panel of the administrator, but not unroll on network new devices with involvement of highly skilled technicians.

For example, in need of expansion of capacity of the Internet gateway, instead of installation of new payments in a basket a blade server, configurations, modifications of tables and so forth, the operator can just appoint new virtual machines from the available pool of resources on which the relevant VNF will be started.

• Speed and efficiency

Contrary to the physical equipment, the network VNF functions can be created and be removed "on the fly", mostly is automated, without the need for attraction of work of technicians.

Such property carries the name "edzhayl" (agile – flexible, operational, quick, effective), the term which took roots in technical literature in the English transcription as it cannot be translated into Russian in a word.

• Scalability and elasticity

Input of new services and application which demand the considerable bandwidth of network in today's conditions often force operators to work in a permanent stress to satisfy everything the growing needs of subscribers and users for new services so that they normally worked at the existing resources. Traditional resources often represent "a bottle neck" (bottleneck) when all other resources of network allow to provide new service without problems, but physical network elements of one-two resources are insufficient, and to expand them – long and it is unprofitable.

This problem is solved in NFV which allows to receive the necessary resources very quickly, unrolling new VNF on VM virtual machines from the available pool of resources.

As these VNF are not limited to parameters of the specialized physical equipment, they can provide property of "elasticity", i.e., they can be unrolled when they are necessary, and are contracted when they are not necessary.

Besides, it allows to avoid a normal situation in traditional network when some network elements are overloaded, and others are underloaded. Fast deployment of VNF allows to distribute the loading which is available at present evenly.

Figure 7. Elasticity of NFV.

• Use of standard IT means

As NFV uses the same infrastructure, as standard data centers, she can use all acceptances of deployment and management acquired in them. It gives the chance to use the existing methods and means of IT for telecommunication networks.

• Fast deployment and disposal of a vendorozavisimost

Because NFV provides means of fast deployment of standard decisions without the excessive costs connected with "monovendor" solutions, operators can get rid of so-called "vendorozavisimost", i.e., excessive attachment to specialized solutions of a small amount of vendors on network.

New solutions can be deployed on network quickly, without the need for waiting of development of new features from traditional vendors that often occupies a progressive tense.

Besides, it allows to unroll quickly new solutions on test domains, to test them, and in case of success of tests, also quickly to unroll them on live network. In case of failure, their cost is minimum as such tests are connected only with installation and start of the software, but not with acquisition and start of the new expensive equipment.

• Simplification of service and technical operation

Service and support of transactions on through NFV allows to lower the possible periods of unavailability of services. For example, failure of the virtual machine on which network function works will immediately cause start of the reserve virtual machine which will execute VNF precisely from the same place on which there was a failure of active VM.

It allows to reach also also modification of the software in the course of work of 24/7 ISSU (In-Service-Software-Upgrade). All this considerably reduces and even completely eliminates the losses connected with faults on network.

• Reduction of capital and operating costs

As the equipment of COTS – inexpensive, in comparison with specialized solutions of traditional telecom vendors, and NFV allows to reach more optimal loading of the equipment, capital costs of operators of creation and development of networks can be considerably reduced.

Reduction of operating costs in NFV manage to be reached due to automation of transactions, increase in a ratio of quantity of items of equipment on one technical specialist and also elimination of need of contents for the staff of specially trained personnel for service of this or that proprietary solution of traditional vendors.

***

These are only some of advantages of application of NFV on telecommunication networks. Many advantages are a consequence from listed and also from their synergy.

Methods of virtualization of network functions as of 2019 are apprehended by chief operators of communication of the world, such as AT&T, Verizon, Nelefonca, Vodafone, China Mobile and many others.

For example, AT&T, with success implements the large-scale project of digital transformation of business which basis is NFV. The AT&T company set the object to virtualize 75% of network functions by 2020 and successfully moves to this purpose.

Volume and dynamics of the market of NFV

2017: data of IDC

In August, 2018 the International Data Corporation (IDC) analytical company provided results of a research of the world market of infrastructure for virtualization of network functions (NFVI). Expenses on the corresponding equipment and the software in 2017 reached $564 million. It is expected that by 2022 revenue in the market will grow for 58.1% annually and will make $5.6 billion.

According to experts, the telecommunication operator it is difficult to transform infrastructure of the networks to create new services, to receive additional sources of income and to improve operational efficiency. In these conditions operators use software-defined networks (SDN) and unroll network functions in the virtualized form factors. Modern telecommunication networks contain both physical, and virtual infrastructure, using the solutions NFV and SDN. However by August, 2018 the market of infrastructure for virtualization of network functions only arises and prepares for rapid growth, the report says.

The market of infrastructure for virtualization of network functions for 2017 was estimated at $564 million

In IDC, using (ETSI) of determination of a network architecture offered by the European institute of standards of telecommunications, select five main segments in the market of NFVI: software-defined calculations, network interaction, storage systems, management and orchestration.

According to forecasts of analysts, orchestration will be the most fast-growing segment thanks to the fact that demand for such services grows for the organization of huge variety of network functions — both physical, and virtual — in several vertical domains.

Wireless infrastructure makes the biggest contribution in the market of NFVI. Routing follows, and the 5G network was become the main driver of growth of expenses in wireless infrastructure of NFVI.^[1]

Reconsideration of the principles of creation of networks

Read Also

• Software-defined networks (Software-Defined Network, SDN) and software-defined DPCs (Software-Defined Data Center, SDDC)
• СRAN - Cloud RAN, Centralized RAN, cloud-based RAN

Notes