Vous êtes sur la page 1sur 17

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

NFV/SDN-based Mobile Networks


Architectures for IoT Services & Applications
Alexander Galvis, Juan F. Botero and Lus A. Fletscher, Members, IEEE

supports and justifies its development. In fact, Gartner's 2014


Abstract Internet of Things (IoT) enables natural Hype Cycle for Emerging Technologies locates IoT in the
communication between multiple devices over a ubiquitous peak of inflated expectations [1] but industry is aware that
platform opening new opportunities for innovation in different with the currently deployed technologies it will be impossible
fields. This survey paper addresses several technological to overcome the challenges imposed by IoT.
challenges faced by the architecture that will support the new IoT The basic idea of IoT is the pervasive presence around us of
services and applications in a mobile 5G environment,
highlighting the key role that Software Defined Networking
a variety of things or objects (such as Radio-Frequency
(SDN) and Network Functions Virtualization (NFV) could play in IDentification (RFID) tags, sensors, actuators, mobile phones,
this evolutive environment. On this emerging and hot topic, this cars, transport infrastructure, appliances, etc.), which, through
paper presents a large review of the existing work and the main unique addressing schemes, are able to interact with each
expected challenges from a perspective of both academy and other and cooperate with their neighbors to reach common
industry. goals [2]. In that way, IoT is an ecosystem of physical objects
capable to share information and even make decisions without
Index Terms Internet of things, network function human intervention in an environment where enabling
virtualization, 5G, mobile network, software defined networking. technologies evolution is driven by applications desirable to
users/stakeholders and a viable marketplace with specific
I. INTRODUCTION needs and requirements [3]. The roadmap shown in Fig. 1

T HIS section briefly describes the three main driven


technologies supporting the content of this survey paper
and the relationship among them: IoT, SDN/NFV and 5G
describes how the IoT ecosystem is likely to evolve in the
coming years and what will be needed to achieve IoT
Roadmap goals.
Mobile Networks. This general scope includes a vision from
academia and industry with emphasis in drivers, enablers,
benefits and strategies to merge these concepts, paradigms and
technologies. Actually, each of these concepts can be analyzed
separately and implemented independently of the others, but
the main goal here is to show how and why integrating them
into a single architecture empowers network operators and
service providers both in the technical and commercial field.
Section I is an introduction to the four technologies
mentioned (IoT, SDN, NFV and 5G). The convergence of
SDN/NFV and modern mobile networks is discussed in
Section II. In turn, Section III discusses the roll of SDN/NFV
in the definition of architectures for IoT applications and
services and how these four technologies work together to
create an efficient, cost effective and scalable environment for
IoT concept implementation. Finally, Section IV summarizes Fig. 1. IoT Roadmap [2]
some conclusions on the subject treated.
A. Internet of Things (IoT) In [2], authors describe the IoT paradigm as a result of the
convergence of different visions, which are explained in
IoT is an idea that evolved into a concept and paradigm detail. Application domains and relevant major scenarios are
over the past 16 years. Today, IoT is a hot topic because the presented too as a vision of the first real evolution of the
technology that enables the interconnection of things in the Internet, giving some example testbeds for each smart
Internet could be currently available and there is a market that environment application domain identified. In addition, a
conceptual IoT framework with Cloud Computing at the
This paper was submitted on March 15, 2017, to the IEEE Transactions on
Network and Service Management, special issue on Advances in Management
center is presented and a model of end-to-end interaction
of Softwarized Networks. between various stakeholders in Cloud centric IoT framework
A. Galvis, J. F. Botero and L. A. Fletscher are with the Engineering is proposed to overcome the complexity related to developing
Faculty, University of Antioquia, Medellin, Colombia. ({alexander.galvisq ; IoT applications using low-level Cloud programming models
juanf.botero ; luis.fletscher}@udea.edu.co.)
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

and interfaces, reducing the time and cost involved in and marketing. A regulatory framework for addressing the
engineering IoT applications [2]. challenge of balancing power, rights and interests in the IoT is
Standardization and normalization processes are also taking also presented.
place within the ITU [4], IETF [5], the European Union [6,7]
B. Network Functions Virtualization (NFV)
and other organizations and institutions worldwide.
Information and descriptions about new horizons for the IoT, Virtualization is not a new technology or trend. Many years
current research, innovation, deployments of IoT, application ago, the idea of virtualize things was applied in computing and
scenarios, pilots, security and privacy issues, current informatics, but the concept of virtualization have been
challenges, open issues, trends and real state of the IoT extended to network elements, and now, to network functions.
ecosystem can be found in [8]. A review of the evolution of the network virtualization
Over the years, with the development of new technologies concept is presented in [12], where a description of Virtual
and market evolution, that vision of the IoT ecosystem has Local Area Networks (VLANs), Virtual Private Networks
expanded considerably. Application of IoT in different (VPNs), active/programmable networks and overlay networks
scenarios will have deep implications, not only regarding how is provided as introduction to Network Virtualization
new technology should be developed, but also in how value Environment (NVE) and its components (business model and
chains and business models will be adapted to the ecosystem architecture). A short review of main projects on this topic
to ensure sustainability and profitability while the user needs until 2009 can also be found in [12], identifying key issues
are met. and open research challenges at that date. The NVE shown in
Nine IoT service sectors are identified in [3]: buildings, Fig. 3 propound principles of coexistence, recursion,
energy, consumer & home, healthcare & life science, inheritance and revisitation. Design criteria for Network
industrial, transportation, retail, security/public safety and IT Virtualization are defined: Flexibility, manageability,
& networks. However, as Gartner shows in [1], IoT is still scalability, isolation, stability and convergence,
maturing and there are a number of factors that limit its programmability, heterogeneity and legacy support.
development and implementation. Most relevant factors are
identified in [3]. No accelerated use and further development
of IoT reference architectures like Architecture Reference
Model (ARM) of the project IoT-A [10], its one of the most
limiting factors for the development of IoT. Many
architectural models have been proposed and each involves
the application of specific technologies, but all of them start
from the same basic model presented in Fig. 2, which
basically describes the vision of the industry [11].

Fig. 3. Network Virtualization Environment [12]

The ultimate vision of virtualization allows for virtualized


network functions (VNFs), which run in the cloud to
maximize efficiency and minimize cost. The NFV concept
appears by first time in October 2012 in a white paper
published by the Network Functions Virtualization Industry
Specification Group (NFV-ISG) at ETSI, which says that NFV
aims to address problems by leveraging standard IT
virtualization technology to consolidate many network
equipment types onto industry standard high volume servers,
switches and storage, which could be located in datacenters,
network nodes and in the end user premises [13]. This means
Fig. 2. Functional (above) and product (below) views of the emerging IoT that practically every function performed in the network can
architecture [11] be implemented by NFV (switching, routing, remote access
servers, firewalls, deep packet inspection, baseband processing
The three core building blocks form the technology
and scheduling, among many others). Fig. 4 depicts the
backbone of the IoT and enable Everything as a Service
evolving NFV architecture [14].
(XaaS): sensors and actuators, embedded processing,
The proposal for Virtual Network Functions (VNFs) is to
connectivity and the cloud are examined in [11]. In addition,
run on Commercial Off-The-Shelf (COTS) hardware. The
authors discuss the impact of IoT and analyze the options for
virtualization layer shown in Fig. 4 is a hypervisor whose task
telcos in the value chain for IoT solutions, taking into account
is to provide virtual access to underlying physical resources,
that a solution is valuable to the end consumer if it is useful,
enabling features like fast start/stop of virtual machines
usable and desirable, a perfect fusion of design, engineering
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

(VMs), snapshot, and VM migration [13]. CPU, RAM, and programmable and the underlying infrastructure to be
storage are flexibly allocated to each VM via software. abstracted for applications and network services. [18]. The
first steps in the development of which would lead to SDN
occurred in 1997 with the GeoPlex by AT&T, but the SDN
concept as it is known today was developed in 2008 in a joint
effort of Stanford University and UC Berkeley. In 2011, the
ONF was created to promote SDN and OpenFlow.
References [19-26] include studies that offer comprehensive
information on SDN, from history to details of the most
remarkable and current developments in the different areas
involved. In particular, [23] provides an almost exhaustive
review of the concepts and technologies into SDN, including
references to specific works in each area, identifying key
issues, highlighting open research problems, describing the
benefits that SDN offers and discussing trends and future work
on the subject.
In the traditional approach, shown in Fig. 5, the
interconnection equipment (switches and routers) performs
both control and data forwarding functions with several
Fig. 4. Simplified NFV architectural framework [14]
implications. One is that the reconfiguration of the control
processes involves tasks on each network equipment, which is
With NFV, operators gain greater flexibility in defining complex and costly. Additionally, protocols and policies
service chains such that a particular service follows the path should be standardized to ensure interoperability between
most applicable to it. Benefits of NFV include lower capital network elements, limiting the options to define the criteria.
expenditures (CAPEX), lower operational expenditures SDN virtue of separating control plane from data plane seems
(OPEX), increased flexibility, scalability and elasticity, and simple but actually it opens a wide range of possibilities for
greater service agility and programmability (if SDN is both services implementation such as network management.
implemented too) [15].
In [16], an industry view of NFV for 2015 can be found,
which provide information about standardization processes
and market forecasts. Authors in [14] describe the NFV
concept and architecture, highlighting related work in ETSI,
Proofs-of-Concept (PoCs) and technical requirements. Two
use cases (virtualization of mobile core network and
virtualization of home network) are presented with their
benefits and challenges, then, open research issues and future
directions of NFV are identified. In turn, [17] discusses
transition from present to future NFV/SDN-based
architectures and networks remarking the transformation of
Service Provider OPEX in both qualitative and quantitative
way.
Fig. 5. Traditional vs SDN-based network [21]
C. Software Defined Networking (SDN)
One of the most interesting disruptive technologies in According the ONF, the main computing trends that are
todays telecommunications landscape is SDN. Something that driving the network change include the changing traffic
came as a proposal to let researchers and network engineers patterns, the consumerization of Information Technology (IT),
perform tests on network infrastructure without affecting the the rise of cloud services, the fact that big data means more
current network configurations, has evolved to become a bandwidth, complexity that leads the stasis, inability to scale
promising approach to solve many of the problems of current and vendor dependence [18]. On the other hand, some of
architectures, meet the challenges imposed by paradigms like benefits of SDN are: enhancing configuration, improving
IoT and achieve the goals of technologies like 5G, among performance, encouraging innovation [25], existence of
other things. evolvable devices, hardware reuse, application awareness,
The Open Networking Foundation (ONF), a user-driven separation of policy and mechanism and higher levels of
organization dedicated to the promotion and adoption of SDN abstraction, isolation and traffic control, centralized network
through open standards development, defines SDN as ...an provisioning and management, independent programmability
emerging architecture that is dynamic, manageable, cost- from standard protocols/policies and more room for realistic
effective, and adaptable, making it ideal for the high- research/experimentation [26]. All these benefits of SDN are
bandwidth, dynamic nature of today's applications. This derived from how its architecture has ben designed and
architecture decouples the network control and forwarding defined. The SDN architecture showed in Fig. 6 is directly
functions enabling the network control to become directly
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

programmable, agile, centrally managed, programmable and, event notification. One value of SDN lies in the expectation
ideally, open standards-based and vendor-neutral [18]. that the Southbound Interface is implemented in an open,
vendor-neutral and interoperable way [18]. One example of a
Southbound Interface is OpenFlow [34].
The Forwarding Devices are logical-physical network
devices that expose visibility and uncontended control over its
advertised forwarding and data processing capabilities. The
logical representation may encompass all or a subset of the
physical substrate resources. The engines and functions may
include simple forwarding between the external interfaces or
internal traffic processing or termination functions [18].
The whole SDN architecture is presented in [18, 20] where
its components are described in detail. In addition, authors in
[23] refer a lot of work related to each single part of the
architecture. These issues include terminology, definitions,
standardization, forwarding devices, southbound interfaces,
network hypervisors, controllers, northbound interfaces,
eastbound/westbound interfaces, language-based
virtualization, programming languages, network applications
and cross-layer issues (debugging, testing and simulation).
Ultimately, in [19], authors identify and explain the SDN
challenges, which are reliability, low level interfaces, ASIC
Fig. 6. Simplified SDN architecture [23] and CPU limitations, interoperability, security, controller
placement, performance and scalability.
Network Applications (SDN Applications) are programs
that explicitly, directly, and programmatically communicate D. 5G Mobile Networks
their network requirements and desired network behavior to Fifth Generation of Mobile Networks (5G) is a concept
the SDN Controller via a northbound interface (NBI). In really new but evolves quite fast, primarily driven by the
addition they may consume an abstracted view of the network industry. 5G, as an evolution of 4G, is more general in scope
for their internal decision making purposes [18]. Examples of because it must provide native support to new concepts such
Network Applications are OpenQoS (traffic engineering) [27], as IoT, Small Cells/Heterogeneous Networks (HetNets) and
OpenRAN (mobile networking) [28] and CloudNaaS the exponential increase of connected devices. In general,
(datacenter networking) [29]. studies predict that by 2020 there will be between 50 and 100
Northbound Interfaces are interfaces between Network billion connected devices, approximately 9.2 billion will be
Applications and SDN Controllers and typically provide mobile terminals and others will come from applications as
abstract network views and enable direct expression of Machine-to-Machine (M2M) and IoT [35]. It is expected that
network behavior and requirements. This may occur at any the new capabilities developed by 5G systems extend the
level of abstraction (latitude) and across different sets of range of applications offered for the deployment of massive
functionality (longitude). One value of SDN lies in the Machine-Type Communications (MTC). The benefits of using
expectation that these interfaces are implemented in an open, M2M cellular connections for MTC include stable
vendor-neutral and interoperable way [18]. There is no performance, security mechanisms and global reach [35].
standardization for this interface [20]. Organizations around the world have begun to work on the
The SDN Controller is a logically centralized entity in definitions and requirements of 5G as well as in the
charge of translating the requirements from the SDN architecture and technologies that will enable its
Application layer down to the SDN Data layer and providing implementation. Those organizations include ITU, ETSI,
the Network Applications with an abstract view of the network ARIB, 4G Americas, WRF, NGMN Alliance, EU, GSMA,
(which may include statistics and events). Definition as a among others [36-40]. In [42], 4G Americas offers a one-time
logically centralized entity neither prescribes nor precludes overview of global 5G initiatives as of the first quarter of
implementation details such as the federation of multiple 2014.
controllers, the hierarchical connection of controllers, Current and future services and environments will impose
communication interfaces between controllers, nor important challenges to 5G systems. To achieve the goals in
virtualization or slicing of network resources [18]. Examples all directions, future cellular networks must be a combination
of SDN Controller or Network Operatong Systems (NOCs) of features and capabilities provided by a special set of
are POX (centralized) [30] and OpenDaylight (distributed) technologies to meet the requirements, as summarized in
[31]. Examples of a virtualization solution and a programming Table 1. Some of the technologies in this table have already
language are FlowVisor [32] and Pyretic [33] respectively. been implemented in existing networks, but a high level of
The Southbound Interface is the interface defined between optimization implementing them in 5G is expected.
an SDN Controller and a Forwarding Devices layer, which
provides at least programmatic control of all forwarding
operations, capabilities advertisement, statistics reporting, and
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

5G expectations and Improvement


Trends/proposals
Several architectures have been proposed but all them are
features dimension
based on the same concepts and technologies, so an interesting
Capacity and
throughput approach to a generic architecture for 5G is shown in Fig. 8 as
improvement, high
Spectrum reuse and use of different ban
presented in [40] where, in addition, use case categories
data rate (~1000x of
throughput Network
(e.g., mm-wave communications using definition and 5G use case families and related examples are
28 ~38 GHz bands), multi-tier network, introduced in an interesting way. NGMN Alliance defines a
improvement over capabilities
D2D communication, C-RAN, massive-
4G, cell data rate native SDN/NFV-based architecture with all the benefits it
MIMO
~10Gbps, signaling
loads less than 1 brings (hardware and software separation, control plane and
~100%) forwarding plane separation and programmability), covering
Reduce latency (2 to
5 ms end-to-end
Network Full-duplex communication, C-RAN, aspects ranging from devices, whole infrastructure, network
capabilities D2D communication
latencies) functions, value enabling capabilities and all the management
Network densification functions to orchestrate the 5G system. APIs are provided on
(~1000x higher
mobile data per unit Network the relevant reference points to support multiple use cases,
Heterogeneous and multi-tier network
area, 100 ~1000x capabilities value creation and business models [40].
higher number of
connecting devices)
Advanced services
and applications (e.g., Enabler for C-RAN, network virtualization, M2M
smart city, service- operational communication, Multi-RAT
oriented sustainability convergence
communication)
Improved energy
Enabler for
efficiency (~10x
operational Wireless charging, energy harvesting
prolonged battery
sustainability
life)
Autonomous
M2M communication, self-organizing
applications and Enabler for
and cognitive network, Multi-RAT
network management, business agility
convergence
Internet of Things
Enabler for
operational
SDN/NFV-based architectures, robust
Reliability (99,999%) sustainability /
fronthaul and backhaul Fig. 8. 5G generic architecture
Enabler for
business agility
Service deployment Enabler for
NFV, orchestration and management
Levels and components in the NGMN 5G architecture are
time (90 min) business agility as follows [40]:
Mobility (~500 Network
Km/h) capabilities
Robust air interfaces, cognitive radio 1. Infrastructure resource layer: physical resources of a
Table. 1. Features, trends, capabilities, requirements and technologies of 5G fixed-mobile converged network.
2. Business enablement layer: library of all functions
Fig. 7 depicts a generic 5G environment [43], required within a converged network in the form of
corresponding to a multi-tier cellular network where several modular architecture building blocks.
enabling technologies can be seen. In some areas, the 3. Business application layer: specific applications and
traditional micro and macro cellular deployment will be used services of the operator, enterprise, verticals or third
to support coexistence/compatibility/interoperability with parties that utilize the 5G network.
legacy infrastructure or due to coverage and capacity 4. E2E management and orchestration entity: contact point
restrictions. to translate the use cases and business models into actual
network functions and slices.
5. 5G Radio Access Technology (RAT) family: set of one or
more standardized 5G RATs that together support NGMN
5G requirements.
6. 5G RAT: component radio interface of the 5G RAT
family.
7. 5G Device: equipment used to connect to a 5G network to
obtain a communication service.
8. 5G System: communications system comprising a 5G
network and 5G devices.

Fig. 9 uses an interesting way to present most of the


concepts that have been discussed about 5G. The end-to-end
5G ecosystem should be architected to meet the expected
demand in 2020 and beyond, so that objectives compliance
will depend on how all these technologies and concepts will
be complemented in the future, and no doubt the industry will
play a very important role in this development. Authors in [43]
offer a summary of emerging technologies for 5G cellular
Fig. 7. 5G cellular wireless network [42] networks with a short description of each one. Meanwhile,
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

[44] presents examples of EU Projects researching technology activities and standardization processes to define new
beyond 4G, a vision of 5G systems with descriptions of architectures for 5G based on SDN/NFV. Comparative
technologies like Ultra-Dense Radio Access Networks analysis among several proposals will be performed
(UDRANETs), modulation techniques, offloading alternatives, highlighting the impact of SDN/NFV implementations in
cognitive radio (CR), software defined radio (SDR), SDN and different network areas and related activities/procedures. In
spectrum use issues. In [45], a description of promising key addition, some available commercial solutions will be
5G wireless technologies is presented with some remarks on presented joint with current industry proposals for medium
future challenges in 5G wireless communication networks. and long term.
Authors in [46] establish an interesting relationship between
challenges, enablers to address these challenges and 5G design A. The Path to SDN/NFV-based Architecture
principles. In [47], Osseiran describes five scenarios for 5G in In cellular wireless networks, both OPEX and CAPEX are
terms of use cases, business models and obstacles. In addition, strongly related to RAN. Reducing these costs is imperative to
a description of five disruptive technologies for 5G can be ensure both profitability and scalability, which is quite
found in [48]. complicated considering that the trend is to systematically
increase the density of transmission sites to achieve the
coverage and capacity goals set by each technology
generation. Although 5G network architecture will be
NFV/SDN-based in a native way, some proposals and
commercial solutions are beginning to apply these concepts in
current 4G networks, specifically in the LTE (Long Term
Evolution) standard and doing a short review of current
commercial solutions, the progress in implementation would
be like Fig. 11 shows.

Fig. 9. A preliminary view of an End-to-End 5G ecosystm [36]

By last, it is clearly expected that the core of the 5G


standardization related to mobile technologies will happen in
the context of 3GPP. However, many standardization activities
related to technologies that will be a part of 5G systems are
being performed in different standardization bodies (IETF,
ETSI, ONF, Open Daylight, OPNFV, Open Stack, ). A
high-level overview of the 5G roadmap is depicted in Fig. 10. Fig. 11. Expected commercial evolution of NFV/SDN implementation in
mobile networks

In the generic architecture of 4G cellular networks, the core


network is basically an all-IP network because there is not any
circuit-switched-based element. That simplifies the abstraction
of the network as a set of functions that can be perform on
generic hardware (virtualization premise) and the decoupling
of data and control planes (software defined networking
premise) can be applied as in a regular data network.
Therefore, the application of NFV/SDN in the core network is
the first that appears naturally, even in 3G/4G networks.
The RAN has a nature quite different from the core
network. Many of the functions performed in the RAN are not
Fig. 10. 5G Roadmap [38] typical of conventional data networks and its distributed
nature itself makes the application of NFV/SDN a more
II. SDN/NFV FOR 5G MOBILE NETWORKS thoughtful issue. Because of this, most of the proposals for
implementing NFV/SDN in mobile networks focus on one of
Further NFV opportunities and network programmability two network areas (RAN or Core) and in some cases they are
are enabled through SDN as a natural solution for 5G cellular oriented to very specific functional aspects such as mobility,
networks, even modifying the mechanisms for services security or scheduling. However, it is likely that the
creation, implementation and management. Therefore, this application of NFV/SDN in the access network brings more
section will describe the efforts and results of research benefits than in the Core. In addition, the control plane in
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

cellular networks is far more complex than in wired networks; these technologies after the PoCs. A well-defined route for
therefore, there will be higher benefits through the flexibility RAN virtualization is discussed in [53] in which C-RAN is
provided by both concepts operating synergistically. On the viewed as an intermediate step towards a complete NFV/SDN-
one hand, NFV can improve the efficiency and flexibility of based mobile network. However, the frontier between RAN
SDN control plane services in both the RAN and Core; SDN and CN is diffuse in a NFV/SDN architecture, but this issue is
can ensure delivery and quality of network traffic between explained in detail in [54] arguing that once the NFV/SDN
virtualized functions of NFV in both network areas. Many implementation phase has been overcome in mobile networks,
high-level applications and services will be optimized by these the concepts of CN and RAN become simple architectural
new features, including IoT services. views of the network rather than architectures itself. Some
related issues and experiences are discussed in [55] [56].
Gartner includes NFV, SDN, IoT and another related
technologies (SmallCells, HetNets, C-RAN) mainly in three of B. State of the Art: The Proposals
their 2016 Hype Cycles: Telecommunications Industry [49], State of the art on SDN, NFV, 5G and IoT is defined by
Software as a Service (SaaS) [50] and Wireless Network work in three related fronts: academic research groups,
Infrastructure [51]. RCR Wireless published a short report one industry and standardization bodies. Table 2 resumes main
year ago analyzing early implementation of NFV/SDN in projects related to NFV/SDN concepts applied to mobile
mobile networks [52]. Last year, the three technologies (NFV, networks.
SDN and IoT) were part of PoCs as well as 5G. All this is
evidence of the great expectation on the implementation of
Backwards
Ref. Solution Network area Concept Short description Year
compatibility
Reprogrammable solution with SDN hierarchical controllers. Network slicing and scheduling are
[57] Double controller SDN Core & RAN SDN Not discussed 2014
performed.
[58] All-SDN Core & RAN SDN Not discussed Two SDN controllers (RAN and CN) with distributed functions. 2015
[61] Smart SDN Core & RAN SDN Not discussed Centralized SDN control with resource abstraction and APIs to provide network intelligence. 2015
[62] SDN in FNA RAN SDN Not discussed FNA-based structure with one SDN controller. Network slicing is performed. 2014
[63] SoftAir Core & RAN SDN No Data and control plane elements are defined by software. Centralized SDN and programmability. 2015
Radio access control in HetNets with network intelligence. Network behavior abstraction using
[64] SoftMobile RAN SDN Yes 2014
layers, centralized SDN.
Functions distribution among BBs and SDN controller. Centralized management specially designed
[65] HSA-SDN RAN SDN Not discussed 2016
for HetNets.
[66] SDN with AHM RAN SDN Not discussed SDN controller with Authentication Handover Module (AHM). Privacy protection. 2015
A single SDN controller that can be implemented in the cloud. Handover decisions are made
[67] SDN centralized RAN SDN No 2014
according to the global view of the network.
Several SDN controllers distributed by geographical areas. Both internal and inter-domain
[67] SDN semi-centralized RAN SDN No 2014
handovers are considered but there isnt a global view of the network.
Multiple layers of controllers: lower layer with distributed SDN controller and upper layers with
[67] Hyerarchical SDN RAN SDN No 2014
centralized SDN controller. There is a global view f the network in a domain scheme.
Centralized SDN controller. A software defined MobileFlow Forwarding Engine (MFFE) for the
[68] MobileFlow Core SDMN Yes 2013
whole network.
Distributed network functions or distributed at the edge. Control plane with three logical controllers.
[69] Plastic SDN RAN SDN Yes 2014
All-SDN cloud architecture.
Backhaul &
[70] Distributed security SDN Yes Centralized EEPC. Data flows through the aggregation level closer to BS and EPC. 2014
Core
Wireless Mesh
[71] Backhaul SDN Yes Centralized SDN for WMN backhaul with SDN transport network from BSs to the EPC. 2015
Network
Mobile traffic offloading to WiFi using SDN in a hierarchical way. Centralized SDN controller plus
[72] Offloading with SDN Backhaul SDN Not discussed 2014
another local controller.
Mobile version of SDN concept with OpenFlow coltrol, network slicing using FlowVisor and NOX
[74] OpenRoad Core SDN Not discussed 2010
as network operator system. SNMP is used and applies for WiMAX and WiFi.
Centralized SDN controller for the whole network and packet classification is moved to the network
[75] SoftCell Core SDN Yes 2013
edge.
Whole RAN abstraction in a huege single and virtual BTS, which performs resorce allocation,
[76] SoftRAN RAN SDN Not discussed
mobility management, load balancing and other control functions in a centralized way.
A virtual EPC for each service and NFV en in the control plane. Hybrid architecture with virtualized
[77] NFV in EPC Core NFV Not discussed 2014
S-GW and P-GW.
[78] vEPC grouping Core NFV Yes Grouping of vEPC entities with distributed functionalities in four interconnected segments. 2014
[79] IPsec Core NFV Yes IP security and NFV environment consolidation. 2015
[80] Access Cloud RAN & Core SDN/NFV No SDN-based architecture with EPC funtions completely virtualized. OpenFlow is used as southbound. 2015
C-RAN resource allocation orchestrated with NFV. MSP cloud infrastructure and intelligente
[81] CSDN RAN SDN/NFV Yes 2015
service orchestration with SDN in a centralized way.
Distributed SDN architecture in the core network. Single SDN controller and a single NFV
[82] SoftNet RAN & Core SDN/NFV Yes 2015
orchestrator inside a NFVI.
Three clouds: access, control and forwarding. Each cloud is SDN-based and interaction between
[83] 3 clouds RAN SDN/NFV No 2015
clouds is considered.
Centralized EPC as a group of VNFs. OpenFlow is used and two SDN controllers: NFV-SDN
[84] SDN-based MPC Core SDN/NFV Yes 2015
controller and SDN controller for E2E connectivity. Management and orchestration of VNFs.
Software defined core network with dynamic traffic management and functions reconfiguration.
[85] Integrated SDMN Core SDMN/NFV No Backbone abstraction using NFV, centralized management and support to SDR and CR concepts 2016
and technologies.
Table. 2. NFV and/or SDN based proposals for future mobile networks
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

Authors in [57] define an architecture to support mobility, Authentication Handover Module (AHM) [66] are deeply
routing management and Connectivity Management as a explained in the works referenced in Table 2, which classifies
Service (CMaaS) in a 5G environment. Specialized and proposals acordint to network segment involved (RAN, CN or
expensive elements like MME, S-GW, P-GW and PCRF have backhaul) and technology applied (SDN, NFV or both).
not been included in the model, as well as tunneling protocols. Alternatives like MobileFlow [68] give the network high
The model incorporates a set of interconnected hierarchical flexibility in such a way that operators can innovate in their
SDN controllers: main controller, RAN controller, BSs domain without relying on content providers or equipment
controller and even an UE controller. Scheduling is performed manufacturers (specially UEs). The SDMN defined by
on geographic areas with multiple eNBs as network slices. MobileFlow comprises a MobileFlow Forwarding Engine
First step towards a SDMN should be the implementation of (MFFE) and a MobileFlow Controller (MFC) as logical
a single controller similar to how its done in computer centralized entities to support operation and innovation in
networks, but RAN and CN separation enhance network network functionalities.
adaptability to multiple scenarios and requirements (IoT for In addition to NFV/SDN implementations within RAN and
example). Authors in [58] propose a hierarchical deployment. CR, some approaches are focused on using these concepts to
The CN controller operates in programmable centralized way enhance the backhaul. [69] [70] [71] [72] propose SDN-based
using OpenFlow as southbound and network functions like solutions in the backhaul which ranging from all-SDN cloud
load balancing and firewall will be implemented as network architectures to SDN-based wireless mesh networks in the
applications in the higher layers. RAN controllers concentrate transport network, making evident the SDN in the
access functions for a set of eNBs and it would be interconnection networks would be the key to achieving the
implemented in the cloud for RAN flexibility and scalability. required efficiency of future mobile networks. Finaly, three
Most of the reviewed proposals make similar approaches more solutions: OpenRoads [74], SoftCell [75] and SoftRAN
with some specific goals such as energy efficiency [59], [76] are well-defined proposals focused on SDN
support to SDR/CR concepts [60], HetNets dynamic implementation in the CN and RAN whose review is
environment support [61] or a Frameless Network recommended.
Architecture (FNA) design [62]. SoftAir [63] and SoftMobile About NFV contributions to mobile network evolution,
[64] stand out among other proposals. SoftAir defines vendors and operators are focused on the required NFVI for
Software Defined Base Stations (SD-BSs) in the RAN and E2E service building and they are taking time to define and
Software Defined Switches (SD-S) in the CN. The whole decide which functions should be virtualized first and where,
control logic is implemented by software and runs on since there are economic and technical reasons why these
conventional computers and remote data centers. Its functions are not virtualized overnight [73]. NFV allows slices
architecture offers five key properties: dynamic definition by EPC virtualization for each type of service. This
programmability on the way, multiple virtual networks in a is quite important for 5G networks and IoT services because
single SoftAir interface, cooperation, open interfaces in data these services scenarios have been defined as use cases on 5G
and control planes, and network visibility from the centralized and IoT is one of them. Maybe, this is the only way to adjust
controllers. Three interesting approaches to SDN the performance of 5G networks to the different requirements
implmenetation in mobile networks are OpenRAN, SoftCell of each type of service, ensuring flexibility, scalability and
and SoftRAN [64]. OpenRAN was proposed by the same team cost-effectiveness [77].
that invented SDN as a mobile version of it and it includes Proposals like described in [78] and [79] are focused on
basic components like OpenFlow for control, FlowVisor for pure NFV application in the CN to support EPC functionalities
slicing and NOX as network operating system, but air as security and mobility management. Others like [80] [81]
interface configuration is made by SNMP (Simple Network [82] [83] [84] and [85] address the integration of SDN and
Management Protocol) in WiFi and WiMAX environments. NFV in the mobile network. AccessCloud [80] and SoftNet
Softcell defines a single controller to gobern all the CN for [82] stand out because they consider implementation in both
scalability and flexibility. SoftRAN, like OpenRAN, was born CN and RAN. Although the RAN is a difficult area to NFV
in Standford University and defines the abstraction of the application, benefits of base-band processing centralization
entire RAN in a huge virtual BTS, which performs resource would be relief for migration of current networks to 5G.
allocation, mobility, load balancing and other functions for all Neves et al. proposed an advanced network management
cell sites in the RAN. SoftRAN faces several challenges like system to 5G based on SDN/NFV [129]. The proposal is a
backhaul capacity dimensioning and latency in the control product of the project SELFNET. SELFNET aims to design
plane. The evolution of these previous concepts and proposals and implement an autonomic framework to achieve self-
resulted in a partial solution for control in Heterogeneous organizing capabilities in managing network infrastructures
Mobile Network (HMN) environments: SoftMobile [64]. [130]. The system architecture defined by SELFNET is
Architecture of SoftMobile is based on a layered model and composed by five layers (infrastructure, data network, control,
used to support efficient operation of RAN in a highly SON autonomic, and SON access). Each layer has different
programmable environment and all network functions are sub-layers with specific tasks. For example, on top of the
performed as applications with legacy networks compatibility.
Physical Sublayer is the Virtualization Sublayer which
As mentioned above, there are many interesting proposals
provides access to the virtual resources of the Data Center
covering the main issues related to RAN, CN and backhaul in
through an hypervisor. The Virtualization Sublayer represents
current and future mobile networks. Concepts like
Harmonized SDN-enabled Approach (HSA) [65] and the NFVI. Another interesting component is the NFV
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9

Orchestration and Management Layer (OML), transversal to form. Computing resources can be provided by the virtual
the first four layers and responsible for orchestrating and infrastructure available in current Cloud Computing platforms.
managing the functions of the SON Sublayers. Between the This infrastructure is able to cope with the integration of
proposed use cases in 5G networks are self-protection, self- monitoring devices, storage devices, analytics tools,
healing, and self-optimization of resource deployment to visualization platforms and client delivery. However, the
maintain QoS levels. requirement of maintaining smart connectivity with current
networks and performing context-aware computation using
C. Early Implementations
network resources can not be accomplished with current WiFi
SDN and NFV are technologies in current development and and 4G-LTE infrastructure. Computing and networking
definition, however player on industry start to building paradigms must go beyond traditional mobile computing
solutions a couple of years ago. For this survey, more than 30
scenarios that use smart phones and portables, and evolve into
companies were consulted, of which 35% are operators and
connecting everyday existing objects and embedding
65% are infrastructure providers. Service providers shown a
intelligence into our environment [110].
great interest in NFV and main vendors like Ericsson, Nokia,
Alcatel-Lucent and Huawei have begun to adapt and update Several studies converge in the fact that robust and reliable
their solutions for NFV support [86] [87]. Other companies infrastructures are needed to support the main challenges that
that provide software at the operator level like WindRiver, IoT still have to face in order to be fully deployed [2] [110].
6WIND, Qosmos and HP have been working with Intel in Quality of service, security/privacy, addressing and
order to optimize software solucin for maximize networking, among others cannot be provided with the ill-
compatibility with NFV and SDN concepts using COTS. Intel suited architecture of current networks. A next-generation
launched its Data Plane Development Kit (DPDK) and architecture shall be aware of all these open issues and,
schedule the launching of a signal processing development kit therefore, it relies in SDN/NFV to provide the needed
to extend and accelerate NFV/SDN adoption [85]. components to overcome the shortcomings of current mobile
Quantitative results about impact of SDN in CAPEX and architectures.
OPEX shown a reduction of 7,72% and 0,31% respectively
compared with LTE non-SDN networks [89]. A. Are Current Network Infrastructures Prepared for IoT?
Alcatel-Lucent launch its first application of NFV to mobile There are three main requirements that will come from an
networks in 2014, a vEPC solution for LTE and beyond [90] enabled IoT environment that shall be addressed by the
already implemented in more than 80 commercial networks. A network infrastructure [111], namely:
solution for RAN virtualization is offered too, called VRAN 1. Capacity: The implementation of small devices able to
[91] that includes a vBBU and a vRNC. Aricent is a company sense the real world and transmit the obtained data to any
working on C-RAN solutions by introduction of NFV/SDN in place, any time, thanks to the Internet, is now a reality due
order to reduce expenditures and support RAN as a Service to the evolution of computing and communications
(RANaaS) [92] [93]. Giant of IT world, Cisco, has a disciplines. With the expecting growing of such devices,
commercial solutions based on NFV and SDN called Virtual the capacity of processing and transmitting all this
Packet Core (VPC) and eXtensible Network Controller (XNC) information is also expected to grow in the coming years.
which are compatible with 2G, 3G, 4G and WiFi [94] [95]. Therefore, current Internet architectures shall evolve in
Maybe, the most complete NFV/SDN-based solution oriented order to support the foreseen capacity.
to mobile infrastructures is SoftCom by Huawei [96], which is
2. Ubiquity: The continuous grow of small devices in the
a unified ICT infrastructure with NFV/SDN enabled. Juniper
network generates ubiquitous sensing and actuating
Networks, Affirmed Networks and Saguna have joined to
perform a PoC of a complete platform called Mobile Edge capabilities that may produce unprecedented scenarios of
Computing (MEC) [97] [98]. More commercial NFV/SDN- smart interaction between virtual and real worlds. Such
based solutions are already offered by Nokia [99], Coriant- ubiquity must be supported from the network
EXFO [100], Radisys [101], Artesyn [102] [103], Aliostar infrastructure by, for example, decoupling device location
[104], NEC [105], Brocade [106], among others. from device identification (current IP addresses are, at the
Service providers are working in these issues too. Verizon, same time, locators and identifiers).
Telefonica, NTT DoCoMo, SK Telecom, among others, are 3. Scalability: Internet infrastructure is currently ossified and
working with several vendors like Ericsson, Juniper, Big it is not able to handle strict QoS guarantees when it
Switch, Ciena, RedHat and Cisco to build the future mobile comes to emergent services (real time streaming, video on
infrastructure facing 5G requirements. demand, etc.) [112]. What would happen with current
Forecasts shown that enterprise market for NFV/SDN will infrastructure with the expected increase in the number of
grow to 18 billion by 2018 [107] [108] [109]. Internet-enabled devices?. The scalability of the TCP/IP
stack will be seriously challenged. For instance, devices
III. SDN/NFV LOWING THE BARRIER FOR IOT APPLICATIONS behind a Network Address Translation (NAT) server are
AND SERVICES opaque. The scalability of the current DNS service will
The foreseen blast of IoT will result in the generation of also be affected with the huge increase in the number of
enormous amounts of data, which have to be stored, processed device elements and networks.
and presented in a seamless, efficient, and easily interpretable
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 10

Consequently, Future Internet (FI) architectures are being deployed in the real network infrastructure [114].
proposed in order to support the requirements imposed by IoT. Here, the interaction between SDN and IoT is reviewed.
As shown in Fig. 12, there is a bidirectional relationship Two interesting branches of recent research have been found:
between FI and IoT; IoT will not be completely deployed SDN in network operators to support IoT and SDN
without a robust and reliable Internet infrastructure that architectures to control IoT.
support the main IoT requirements, in turn, a key element of
FI infrastructures shall be to support ubiquitous and scalable SDN in network operators to support IoT
IoT applications and its interactions with current Internet
equipment. In [115], the authors describe how SDN can be integrated to
current mobile and cloud infrastructures using two use cases:
i) An SDN switch acting as an IoT gateway controlled by a
cloud-based SDN controller, and ii) an eNodeB where SDN is
used to offload data related traffic from the Evolved Packet
Core (EPC) to improve quality of experience for the users in
LTE networks.
The first use case is proposed for home environments
where IoT is used in smart automated appliances such as,
HVAC, washing/drying machines, smart plugs lighting,
multimedia, etc. All the data generated by these
appliances shall be protected from external attacks,
provided with confidentiality and integrity, and prioritized
Fig. 12. Relationship between FI and IoT with regard to non-critical traffic. All these functionalities
can be currently implemented by expensive specific
Software Defined Networking and Network Functions destination hardware (Firewalls, IPSec/IKE, Traffic
Virtualization are key technologies to enable FI architectures shapers, etc.). However, if the IoT gateways in smart
to support the aforementioned requirements [113]. On the one homes are cheap SDN switches, all the intelligence to
hand, SDN separates the control plane from the data plane; provide the required functions can be concentrated in the
this will increase the flexibility and adaptability of the network SDN controller which will be located at the operator
infrastructure: network programmability provided by SDN premises. Apart from that, the controller performs
makes this infrastructure less dependent on the protocols and, provisioning, configuration, administration and upgrades
therefore, it becomes more scalable. On the other hand, NFV and maintenance of the IoT gateways in an efficient and
allows to decouple hardware from functionality by virtualizing cost-effective way.
the network functions; this allows the operators to use The second proposed use case shows how the introduction
commodity instead of specialized hardware when building of SDN in the current LTE architecture may help in
their infrastructure, resulting in a significant reduction of their reducing the traffic load by introducing traffic
operative and capital expenditures (OPEX and CAPEX). classification. Currently, all traffic (voice and data)
Besides, NFV enables the dynamic provision of network follows the same path before going into the Internet -
functionality, which contributes to the deployment of eNodeB Serving Gateway (S-GW) Packet data
ubiquitous services. network gateway (PDN-GW). By installing a SDN switch
Currently, some research initiatives have started to propose in the eNodeB, policies may be installed that reduce the
the interactions SDN-IoT and NFV-IoT. Next, these two traffic load by routing the traffic that can be offloaded to
interactions are deeply surveyed. the Internet (e,g. video, file transfer or P2P) directly
B. Software Defined Networking Enabling the Internet of without having to traverse the S-GW and the PDN-GW.
Things
In [116], Ojo et al. proposed a general SDN-IoT
As we have previously mentioned, current Internet
architecture with NFV implementation. They studied where
infrastructure was not designed to support the high level of
and how to adopt SDN and NFV according to the challenges
scalability, high amount of traffic and mobility introduced by
of IoT. The main contribution of authors is to analyze the
the birth of IoT. The infrastructure responsible for transmitting
business perspective of IoT by considering SDN and NFV as
the information coming of IoT devices should be adapted to
market enablers.
the new post-PC services, while providing security, stability,
high rate, reliability and availability among others. The recent SDN architectures to control IoT
idea of add programmability to the network in order to provide
customization of the network behavior can overcome some of Several approaches to control IoT using the SDN paradigm
IoT requirements. SDN is able to offer this programmability; has been recently proposed. Some of them tackle Wireless
besides, it offers the flexibility to take decisions on IoT Sensor Networks (WSN), which always had been thought to
networks using customized algorithms that can be instantly be application-specific [117]. However, this belief can lead to
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 11

resource underutilization (multiple WSNs for respective such architecture is to offer Sensing-as-a-Service where a
applications are deployed in the same or overlapping terrain) user request a sensing service that is provided through the
and counterproductivity (different vendors develop WSN in cloud sensing architecture (see Fig. 13a).
isolation). Therefore, recent research proposes a Software-
Defined WSN architecture to face these shortcomings:
In [118], the authors claim that in current WSN
architectures, each node implements all network layers, from
physical to application, and therefore, they act in an
autonomous way making the WSN inelastic and hard to
manage. To overcome these limitations, they propose a
programmable data plane that can be manipulated from a
logically centralized controller via a new protocol called
Sensor OpenFlow (SOF); the idea is to transform traditional
WSN networks in versatile (sensors become application-
customizable), flexible (easier policy enforcement due to
centralization) and easy to manage (open APIs allow to easily
add new applications) networks.
Implementing a SOF protocol is very challenging as the (a)
conditions of a WSN environment are very different from the
ones SDN and OpenFlow are currently made for (wired
environments). Here, the authors propose the following
strategies to cope with the main challenges to build Software
Defined Wireless Sensor Networks:
Creating flows in SDN is typically made based on address
match fields (IP/MAC addresses or MPLS labels), in
contrast, WSN are data-centric networks so the flow
matching is made based on concatenated-attribute-value
pairs, for instance, 30<temperature<60 and Zone-ID=7
(b)
AND x-coordinate>150. Fig. 13. Software-Defined Sensor Networks a) Cloud Sensing Architecture
In SDN, the OpenFlow channel transmits messages b) Software-Defined Sensor Architecture [120]
between a controller and a switch using the TCP/IP stack.
To solve this problem in WSN, the SOF channel is To deploy this cloud-WSN integrated architecture today,
supplied by overlaying a transport protocol directly over multiple application-specific WSNs should be deployed in the
WSN [119]. same area. Here, the authors propose Software-defined sensor
To avoid the congestion overhead introduced by the networks (SDSNs) consisting on programmable sensor nodes
control channel between the controller and each sensor, that can change their functionalities by loading on-demand
SOF just send a flow-setup request to the controller for programs according to the requests performed by the end user.
the first miss in its flow table. The remaining associated SDSNs will be integrated in the cloud-WSN architecture (see
packets are suppressed until the corresponding answer is blue ballons in Fig. 13a), so that if the deployed sensors
received from the controller. cannot satisfy the sensing requirements offered by the cloud,
In traditional SDN, end nodes are out of the scope of on-demand programs can be injected to these sensors to adapt
OpenFlow as it communicates just with forwarding their functionality to the user requests.
switches. In WSN, end nodes act as SDN-switches (they The architecture of the SDSNs is detailed in Fig. 13b. The
forward data) and traffic generators. To cope with this main goal of a SDSN is to deploy sensing tasks that are
difference, a traffic generation module is added to each requested from the Sensing-as-a-Service infrastructure. The
sensor; this module is in charge of reading data from the main roles of the SDSN architecture are:
sensing hardware, and converting the data if needed. Sensor Control Server: It provides a role (e.g. cluster
In contrast to traditional SDN networks, in WSN data is head, data aggregator, temperature sensing, etc.)
sometimes processed in-situ in order to reduce data generation and delivery mechanism, where a scenario
redundancy to optimize bandwidth and energy. To solve compiler generates sensor role programs based on a
this difference, the proposed architecture adds an in-net scenario description that specifies the functions to be
proc module to each sensor that modifies the packet as activated in sensor nodes. This element can be seen as the
needed and passes it to the flow table. counterpart of a controller in a SDN traditional
architecture.
In [120], the authors propose an architecture based on the Sensor Nodes: As in traditional WSNs, they are the end
convergence of WSN and cloud-computing. The main aim of nodes of the architecture in charge of sense the
environment and transmit the useful data to a sink.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 12

However in SDSNs, these nodes include a mechanism to them; this is a challenging and complex task that is solved in
dynamically initiate a role in response to the user the paper by means of network calculus and genetic
requests. By invoking a generated role program algorithms. Finally, 4) the communications layer triggers the
downloaded to a sensor, the sensor can adjust its necessary communications in the IoT Multinetworks (e.g.,
functionalities and properties to perform the route the video data sent from Camera 001 via Ethernet).
corresponding task. The aforementioned SDN-based architectures for IoT have
not accounted for a paramount feature of large-scale (urban)
Although in the article, the authors do not mention how the IoT environments: mobility. To cope with this important
sensor control server would communicate with the end sensor requirement, a new architecture, called UbiFlow: A software-
nodes to perform the network control; it is possible to think defined IoT system for ubiquitous flow control and mobility
that a control channel such as the above-mentioned SOF management in urban heterogeneous networks. The
would perfectly fit in this architecture. architecture of Ubiflow is depicted in Fig. 15 where multiple
Besides from WSN networks, SDN has been proposed as an controllers (each with the internal architecture presented in
enabler technology to propose architectures to manage and Fig. 14) are deployed in different geographical areas (called
control IoT environments. For instance, in [121], the authors partitions).
propose an IoT management architecture where the controller
manages the network topology. It has the advantage that the
IoT network becomes independent of its gateway; that is, if
the gateway breaks down, as a SDN controller manages it, the
network can reconfigure accordingly and the gateway can be
updated.
A wider SDN-based architecture of IoT is presented in
[122]. Here, the architecture is proposed to overcome the
following shortcomings of the heterogeneous multi-networks
IoT nature: 1) shared provisioning of network and sensor
resources across applications, and 2) interoperability challenge
that arises when heterogeneous devices exploit different data
formats for modeling information and diverse protocols for
Fig. 15. UbiFlow System Architecture [123]
machine-to-machine (M2M) data exchange. The proposed
architecture of the SDN-IoT controller is shown in Fig. 14. Thanks to its distributed architecture, UbiFlow has a
consistent and updated state of the multiple-IoT networks, its
key contributions are:
A distributed overlay infrastructure of SDN controllers to
achieve mobility management and fault tolerance in the
multiple IoT networks in an urban environment. All IoT
devices in a single partition request different types of
dataflows that are analyzed and scheduled by the
controller responsible of this partition by means of robust
optimization algorithms. To achieve network consistency,
the controllers form an overlay network where they
interchange meaningful information that allows the whole
infrastructure to provide mobility management (Mobile
Fig. 14. IoT Controller Architecture [122] SDN-based handover) and scalable control.
It is composed of a data collection component that allows An optimal assignment algorithm for the controller that
storing the information of the multiple IoT network allows an efficient schedule of the dataflows requested by
environments and presenting it to a layered set of components. the IoT multiple networks. This algorithm is based in
These components represent the way to abstract the network calculus and optimization heuristics.
heterogeneity of the different IoT network and devices in four A load balancing scheme for distributed controllers by
layers: 1) The task resource matching layer is in charge of analyzing the variations in flow traffic characteristics.
defining high level tasks (e.g. determine how many vehicles C. Network Functions Virtualization Enabling the Internet of
currently there are in a recharging station), each task is then Things
subdivided in several services (e.g. capture video from As NFV is an emerging paradigm, few approaches have
recharging station + count vehicles) managed by 2) the tackled the impact that it may have in the deployment of future
Service and Solution Specification layer. After services have architectures able to support IoT. These approaches define
been defined, 3) the Flow Scheduling layer takes the NFV in the core of the architecture to reduce the OPEX and
requirements of each service and schedule flows to fulfill CAPEX of the operators. Besides, it is always integrated with
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 13

SDN performing a centralized control to take decisions over one generated by humans, 2) IoT requirements against
the virtualized architecture. network, here the challenge is to understand the impact of the
In [124], the authors claim that the architectural evolution upcoming machine-generated traffic in the network
required by an IoT-based Internet shall be based on the architecture (i.e. impact in security, privacy, mobility,
combination of SDN and virtualization (including NFV) as scalability, QoS, resilience, etc.) 3) Architecture requirements
Fig. 16 shows. They propose a new architecture built for IoT, where the challenge is to converge different
according to two virtualization levels: 1) network level, where technologies that support IoT in a common architecture, 4) IoT
virtualization is used to abstract network elements in a infrastructure services deals with the challenging
distributed pool that may involve different physical locations implementation of the Infrastructure as a Service (IaaS)
[125], and 2) end-user level, where virtualization is used to concept to offer this infrastructure to service providers by
improve the connectivity among different IoT user devices means of emergent technologies (e.g. SDN, network
(e.g. sensor virtualization). virtualization and NFV), and 5) IoT overall impact on
architecture, where the challenge is to design an architecture
able to cope with the specifics of IoT traffic (unpredictability,
fluctuation, elasticity, etc.).
This work also claims that NFV + SDN can be very
beneficial for IoT based network environments. SDN will
deploy the services by allocating Virtual Network Functions
(VNF) in the places decided by the NFV orchestrator. In turn,
NFV decouples infrastructure services from resources. Thus,
IoT use-cases can be isolated preventing the risk of impacting
one infrastructure by different services.
To face the aforementioned challenges making use of SDN
and NFV technologies, this approach proposes an SDN/NFV-
based IoT architecture to provide infrastructure service (cf. Fig
17).

Fig. 16. Virtualization levels and layers [124]

The proposed architecture is composed on three main layers


built on top of the aforementioned virtualization levels:
The SDN control layer in charge of providing centralized
control to the network dataplane. This dataplane may
belong to different virtual networks that are instantiated
according to the application demands. It manages, by
means of routing and shaping, the traffic flows traversing
the network devices.
The Thin software virtualization layer is installed in end-
user devices to abstract sensor network subsets that serve
different applications. The main goal of this layer is to
abstract sensor functionalities to allow interoperability
among different vendors. Fig. 17. NFV/SDN-based IoT infrastructure [126]
Finally, the Network infrastructure virtualization layer
offers an abstract view of the core network equipment as The architecture is divided in layers and actors. On the one
virtual networks. It works together with the SDN control hand, the main actors of the IoT ecosystem are: 1) IoT Service
layer to facilitate efficient allocation and usage of Broker that aggregates connected objects and provides links
networking resources. for different application domains, 2) Object Owner that buys
and maintains the objects (things), 3) Object User who uses
In [126], an architecture based in NFV + SDN is proposed the object, 4) Data Owner that stores and manages the data
to overcome the main challenges posed by the IoT arrival. gathered from connected objects, 5) Data Analyst that perform
These paper highlights the following challenges: 1) IoT the analysis of the collected data, 6) Data User that can access
market, services, and use cases taking into account the certain data portion, depending on the services he/she has
explosive growth of IoT due to the fact that in an IoT access to and, finally, 7) B2B Partner that contracts with the
ecosystem the traffic generated by machines may exceed the IoT Service Broker to benefit from infrastructure services. On
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 14

the other hand, the layers of the architecture from top to [5] Isam Ishaq, David Carels, Girum K. Teklemariam, Jeroen Hoebeke,
Floris Van den Abeele, Eli De Poorter, Ingrid Moerman and Piet
bottom are: 1) Service layer that groups all the service-level Demeester. IETF Standardization in the Field of the Internet of Things
functions at high level, 2) Global OS layer that contains the (IoT): A Survey. Journal of Sensor and Actuator Networks, 2013, 2,
orchestration NFV system and the SDN controllers, it decides 235-287
[6] Harald Sundmaeker, Patrick Guillemin, Peter Friess and Sylvie
how to deploy the services in the infrastructure based on the
Woelffl. Vision and Challenges for Realising the Internet of Things.
data analysis and manages the configuration of all network Cluster of European Research Projects on the Internet of Things (CERP-
elements, finally, 3) Virtualization layer in charge of IoT), European Union, 2010.
managing the hardware resources and their relationship with [7] Ovidiu Vermesan and Peter Friess. IoT Governance, Privacy and
Security Issues. Position paper by Cluster of European Research Projects
virtual machines made available to the above layers. on the Internet of Things (CERP- IoT), European Union, 2015.
[8] Ovidiu Vermesan and Peter Friess. Building the Hyperconnected
D. Upcoming SDN/NFV-based IoT network architectures Society: IoT Research and Innovation Value Chains, Ecosystems and
As it have been reviewed in the previous chapter, several Markets. River Publishers, River Publishers Series in Communications,
2015.
approaches for the upcoming 5G architecture based on NFV [9] M2M/IoT Sector Map. Beecham Research Limited.
and SDN are being proposed by the research community. http://www.beechamresearch.com/article.aspx?id=4. Accessed 13 Jun
Even, recent proposals aim to integrate SDN and NFV for 2015.
future network architectures [127] [128]. However, almost [10] Introduction to the Architectural Reference Model for the Internet of
Things. Internet Of Things Architecture (IoT-A) Project, Seventh
none of them specifies how these proposals support the Framework Programme at EU, 2013. (http://www.iot-a.eu/)
foreseen blast of the IoT. [11] Mark Fell, Hanne Melin. Roadmap for the emerging of Internet of
Here, it is considered that the main challenge for the Things. Carr & Strauss, 2014.
creation of an NFV/SDN-based 5G architecture will be to [12] Chowdhury, N.M.M.K.; Boutaba, R., "Network virtualization: state of
the art and research challenges," Communications Magazine, IEEE,
make them suitable to cope with the requirements of the IoT vol.47, no.7, pp.20,26, July 2009
arrival. This challenge can be meet by finding means to adapt [13] Pierre Lynch, Michael Haugh, Liza Kurtz, Joseph Zeto. Demystifying
IoT architectures such as the presented in this chapter, to the NFV in Carrier Networks: A definitive guide to successful migrations.
Ixia, 2014.
5G architecture that is being proposed by the academy and [14] Bo Han; Gopalakrishnan, V.; Lusheng Ji; Seungjoon Lee, "Network
industry nowadays. function virtualization: Challenges and opportunities for innovations,"
Communications Magazine, IEEE, vol.53, no.2, pp.90, 97, Feb. 2015.
[15] Frank Yue. Network Functions Virtualization Everything Old Is New
IV. CONCLUSIONS AND FURTHER WORK Again. White paper by F5 Networks, Inc. 2013.
This section describes the current efforts from both research [16] Dan Meyer. Network Functions Virtualization: Standards, ecosystem
and 2015 predictions. RCR Wireless News Group, Frebruary 2015.
community and industry to propose convergent architectures [17] Hernandez-Valencia, E.; Izzo, S.; Polonsky, B., "How will NFV/SDN
for the future generation of mobile access networks, able to transform service provider opex?," Network, IEEE , vol.29, no.3,
manage the increasing complexity generated by the growing pp.60,67, May-June 2015.
[18] Open Networking Foundation (www.opennetworking.org).
introduction of IoT devices in the coming years. This section
[19] Manar Jammal, Taranpreet Singh, Abdallah Shami, Rasool Asal,
emphasizes the importance of the integration of SDN/NFV in Yiming Li, Software defined networking: State of the art and research
the 5G architecture to handle this complexity that will force challenges, Computer Networks, Volume 72, 29 October 2014, Pages
the network control plane to provide an efficient 74-98, ISSN 1389-1286
[20] Akram Hakiri, Aniruddha Gokhale, Pascal Berthou, Douglas C.
administration of a huge amount of heterogeneous end devices Schmidt, Thierry Gayraud, Software-Defined Networking: Challenges
with different application-specific constraints. The role of and research opportunities for Future Internet, Computer Networks,
NFV here will be to decouple hardware from functionality by Volume 75, Part A, 24 December 2014, Pages 453-471, ISSN 1389-
1286
means of virtualization. In turn, SDN will enforce the service [21] Nunes, B.A.A.; Mendonca, M.; Xuan-Nam Nguyen; Obraczka, K.;
orchestration by allocating the virtual appliances in the most Turletti, T., "A Survey of Software-Defined Networking: Past, Present,
adequate places in the network. It is then shown how IoT and Future of Programmable Networks," Communications Surveys &
Tutorials, IEEE, vol.16, no.3, pp.1617,1634, Third Quarter 2014
applications and services will benefit from all this evolution. [22] Fei Hu; Qi Hao; Ke Bao, "A Survey on Software-Defined Network and
Finally, some conclusions will be presented and raises the OpenFlow: From Concept to Implementation," Communications
main identified challenges. These challenges mainly belong to Surveys & Tutorials, IEEE, vol.16, no.4, pp.2181, 2206, Fourthquarter
2014
the integration of SDN/NFV and 5G to improve the [23] Kreutz, D.; Ramos, F.M.V.; Esteves Verissimo, P.; Esteve Rothenberg,
manageability introduced by IoT that was discussed in Section C.; Azodolmolky, S.; Uhlig, S., "Software-Defined Networking: A
III. Comprehensive Survey," Proceedings of the IEEE, vol.103, no.1,
pp.14,76, Jan. 2015
[24] Hamid Farhady, HyunYong Lee, Akihiro Nakao, Software-Defined
REFERENCES Networking: A survey, Computer Networks, Volume 81, 22 April 2015,
[1] Gartner's 2014 Hype Cycle for Emerging Technologies Maps the Pages 79-95, ISSN 1389-1286
Journey to Digital Business. Gartner, Inc. [25] Wenfeng Xia; Yonggang Wen; Chuan Heng Foh; Niyato, D.; Haiyong
http://www.gartner.com/newsroom/id/2819918. Accessed 13 Jun 2015. Xie, "A Survey on Software-Defined Networking," Communications
[2] Atzori, L., Iera, A., & Morabito, G. (2010). The internet of things: A Surveys & Tutorials, IEEE, vol.17, no.1, pp.27,51, Firstquarter 2015
survey.Computer networks, 54(15), 2787-2805. [26] Macedo, D.F.; Guedes, D.; Vieira, L.F.M.; Vieira, M.A.M.; Nogueira,
[3] Ovidiu Vermesan and Peter Friess. Internet of Things: Converging M., "Programmable Networks From Software-Defined Radio to
technologies for smart environments and integrated ecosystems. River Software-Defined Networking," Communications Surveys & Tutorials,
Publishers, River Publishers Series in Communications, 2013. IEEE, vol.17, no.2, pp.1102, 1125, Secondquarter 2015
[4] The Internet of Things. ITU Internet Reports by International [27] H. Egilmez, S. Dane, K. Bagci, and A. Tekalp, OpenQoS: An Open-
Telecommunications Union (ITU), Geneva, November 2005. Flow controller design for multimedia delivery with end-to-end quality
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 15

of service over software-defined networks, in Proc. Asia-Pacific Signal [55] Monica Paolini. "RAN virtualization to achieve sustainable growth - A
Inf. Process. Assoc. Annu. Summit Conf., 2012, pp. 18. conversation with Javier Lorca Hernando from Telefonica". Senza Fili,
[28] M. Yang et al., OpenRAN: A software-defined ran architecture via August 2015.
virtualization, in Proc. ACM SIGCOMM Conf., 2013, pp. 549550. [56] Monica Paolini. "The emergence of the NFV ecosystem: Laying the
[29] T. Benson, A. Akella, A. Shaikh, and S. Sahu, Cloudnaas: A cloud foundation for a new way to run mobile networks". Senza Fili,
networking platform for enterprise applications, in Proc. 2nd ACM November 2014.
Symp. Cloud Comput., 2011, pp. 8:18:13. [57] R. Trivisonno, R. Guerzoni, I. Vaishnavi and D. Soldani. SDN-based
[30] M. McCauley, POX, 2012. (www.noxrepo.org) 5G mobile networks: architecture, functions, procedures and backward
[31] OpenDaylight, A Linux Foundation Collaborative Project, 2013. compatibility. Transactions on emerging telecommunications
(www.opendaylight.org) technologies, 2014.
[32] R. Sherwood et al., FlowVisor: A network virtualization layer, [58] V. Yazc, C. Kozat, and M. Sunay. A New Control Plane for 5G
Deutsche Telekom Inc. R&D Lab, Stanford, Nicira Networks, Tech. Network Architecture with a Case Study on Unified Handoff, Mobility,
Rep., 2009. and Routing Management. IEEE Communications Magazine,
[33] J. Reich, C. Monsanto, N. Foster, J. Rexford, and D. Walker, Modular November 2014.
SDN programming with pyretic, USENIX Mag., vol. 38, no. 5, Oct. [59] R.L.G. Cavalcante et al., Toward Energy-Efficient 5G Wireless
2013. Communications Technologies: Tools for Decoupling the Scaling of
[34] Nick McKeown, Tom Anderson, Hari Balakrishnan, Guru Parulkar, Networks from the Growth of Operating Power, IEEE Signal Proc.
Larry Peterson, Jennifer Rexford, Scott Shenker, and Jonathan Turner. Mag., vol. 31, no. 6, Nov. 2014, pp. 2434.
2008. OpenFlow: enabling innovation in campus networks. ACM [60] A. Dalvi, P. K. Swamy, and B. B. Meshram, Cognitive Radio:
SIGCOMM Computer Communication Review, Volume 38, Issue 2, Emerging Trend of Next Generation Communication System, Wireless
Pages 69-74, April 2008. Communication, Vehicular Technology, Information Theory and
[35] Ericsson Mobility Report. Ericsson AB, June 2015. Aerospace & Electronic Systems Technology (Wireless VITAE), 2011.
[36] 4G Americas Recommendations on 5G Requirements and Solutions. 4G [61] S. Sun, L. Gong, B. Rong, and K. Lu, An Intelligent SDN Framework
Americas, October 2014. for 5G Heterogeneous Networks. IEEE Communications Magazine
[37] Demestichas, P.; Georgakopoulos, A.; Karvounas, D.; Tsagkaris, K.; November 2015.
Stavroulaki, V.; Jianmin Lu; Chunshan Xiong; Jing Yao, "5G on the [62] X. Xiaodong, Z. Huixin, D. Xun, H. Yanzhao, T. Xiaofeng, Z. Ping.
Horizon: Key Challenges for the Radio-Access Network," Vehicular SDN Based Next Generation Mobile Network With Service Slicing and
Technology Magazine, IEEE, vol.8, no.3, pp.47,53, Sept. 2013 Trials. China Communications February 2014.
[38] 5G Vision: The 5G Infrastructure Public Private Partnership - The next [63] I. Akyildiz, P. Wang, S. Chun-Lin. SoftAir: A software defined
generation of communication networks and services. 5G-PPP supported networking architecture for 5G wireless systems. I.F. Akyildiz et al. /
by the European Commission, February 2015. Computer Networks 85 (2015) 118.
[39] 5G Vision and Requirements. IMT-2020 (5G) Promotion Group, May [64] T. Chen, H. Zhang, X. Chen and O. Tirkkonen. SoftMobile - Control
2014. evolution for future heterogeneous mobile networks. IEEE Wireless
[40] 5G White Paper. Next Generation Mobile Networks Ltd. (NGMN Communications, December 2014.
Alliance), February 2015. [65] A. Akhtar, X. Wang, and L. Hanzo. Synergistic Spectrum Sharing in
[41] Understanding 5G: Perspectives on future technological advancements 5G HetNets: A Harmonized SDN-Enabled Approach. IEEE
in mobile. GSM Association GSMA Intelligence, December 2014. Communications Magazine (Volume:54 , Issue: 1 ), January 2016.
[42] 4G Americas Summary of Global 5G Initiatives. 4G Americas, June [66] X. Duan and X. Wang. Authentication Handover and Privacy
2014. Protection in 5G HetNets Using Software-Defined Networking. IEEE
[43] Hossain, E.; Hasan, M., "5G cellular: Key enabling technologies and Communications Magazine, April 2015.
research challenges," Instrumentation & Measurement Magazine, IEEE , [67] S. Kuklinski, Y. Li, K. Truong. Handover Management in SDN-based
vol.18, no.3, pp.11,21, June 2015 Mobile Networks. Globecom 2014 Workshop - The 6th IEEE
[44] Ral Chvez-Santiago, Micha Szydeko, Adrian Kliks, Fotis Foukalas, International Workshop on Management of Emerging Networks and
Yoram Haddad, Keith E. Nolan, Mark Y. Kelly, Moshe T. Masonta and Services.
Ilangko Balasingham. 5G: The Convergence of Wireless [68] K. Pentikousis, Y. Wang, and W. Hu MobileFlow - Toward Software-
Communications. International Journal on Wireless Personal Defined Mobile Networks. Huawei Technologies. IEEE
Communications, Springer US, March 2015. Communications Magazine, July 2013.
[45] Cheng-Xiang Wang; Haider, F.; Xiqi Gao; Xiao-Hu You; Yang Yang; [69] R. Guerzoni, R. Trivisonno, D. Soldani. SDN-Based Architecture and
Dongfeng Yuan; Aggoune, H.; Haas, H.; Fletcher, S.; Hepsaydir, E., Procedures for 5G Networks. 2014 1st International Conference on 5G
"Cellular architecture and key technologies for 5G wireless for Ubiquitous Connectivity (5GU).
communication networks," Communications Magazine, IEEE , vol.52, [70] V. Jungnickel, K. Habel, M. Parker, S. Walker, C. Bock, J. Ferrer.
no.2, pp.122,130, February 2014 Software-Defined Open Architecture for Front - and Backhaul in 5G
[46] Agyapong, P.; Iwamura, M.; Staehle, D.; Kiess, W.; Benjebbour, A., Mobile Networks. ICTON 2014 We.B4.4.
"Design considerations for a 5G network architecture," Communications [71] K. Seppanen, J. Kilpi, T. Suihko. Integrating WMN Based Mobile
Magazine, IEEE, vol.52, no.11, pp.65,75, Nov. 2014 Backhaul with SDN Control. Springer, Mobile Netw Appl (2015)
[47] Afif Osseiran. Five alive! - 5G beyond the hype. Ericsson Business 20:3239.
Review, Issue 2, 2014. [72] J. Costa, R. Kantola, J. Llorente, V. Ferrer, J. Manner, A. Yi-Ding, Y.
[48] Boccardi, F; Heath, R.W.; Lozano, A.; Marzetta, T.L.; Popovski, P., Liu, S. Tarkoma. Software Defined 5G Mobile Backhaul. 2014 1st
"Five disruptive technology directions for 5G," Communications International Conference on SG for Ubiquitous Connectivity (SGU).
Magazine, IEEE , vol.52, no.2, pp.74,80, February 2014 [73] Signals Research Group (2014), Signals Ahead, Vol. 9, Number 9,
[49] "Hype Cycle for the Telecommunications Industry, 2015". Gartner, Inc. Exploring the world of Wireless SDN and NFV: its not a single
25 July, 2015. network anymore.
[50] "Hype Cycle for Software as a Service, 2015". Gartner, Inc. 29 July, [74] K.-K. Yap et al., OpenRoads: Empowering Research in Mobile
2015. Networks, ACM SIGCOMM Comp. Commun. Rev., vol. 40, no. 1,
[51] "Hype Cycle for Wireless Networking Infrastructure, 2014". Gartner, 2010, pp. 12526.
Inc. 24 July, 2015. [75] X. Jin et al., SoftCell: Scalable and Flexible Cellular Core Network
[52] Dan Meyer. "NFV, SDN and cloud: When, where and how deep?". RCR Architecture, 9th ACM Conf. Emerging Networking Experiments and
Wireless. March, 2016. Technologies, 2013, pp. 16374.
[53] Monica Paolini. "Charting the path to RAN virtualization: C-RAN, [76] A. Gudipati et al., SoftRAN: Software Defined Radio Access
fronthaul and HetNets". Senza Fili, August, 2015. Network, 2nd ACM SIGCOMM Wksp. Hot Topics in Software
[54] Monica Paolini. "Moving past C-RAN, to get the performance and cost Defined Networking, 2013, pp. 2530.
gains of full RAN virtualization - A conversation with Gilad Garon from [77] Bringing network function virtualization to LTE. 4G Americas.
ASOCS". Senza Fili, July 2015. November 2014.
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 16

[78] H. Hawilo, A. Shami, M. Mirahmadi, and R. Asal. NFV: State of the http://www.itworld.com/article/2694257/networking/sdn-market-could-
Art, Challenges, and Implementation in Next Generation Mobile hit--18b-by-2018.html
Networks (vEPC). IEEE Network, November/December 2014. [108] How big is the SDN Market?. SDx central, 2013.
[79] K. Faheem, Dr K. Rafique. Securing 4G/5G wireless networks. May https://www.sdxcentral.com/reports/sdn-market-size-infographic-2013/
2015, Computer Fraud & Security. [109] M. Howard. NFV Market to Grow More than 5-Fold Through 2019,
[80] P. Ameigeiras, J. Ramos, L. Schumacher, J. Prados. Jorge Navarro- Says HIS. Infonetics Research, 2016.
Ortiz, and Juan M. Lopez-Soler Link-Level Access Cloud Architecture http://www.infonetics.com/pr/2015/NFV-Market-Highlights.asp
Design Based on SDN for 5G Networks. IEEE Network, March/April [110] Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of
2015. Things (IoT): A vision, architectural elements, and future directions.
[81] A. Bradai, K. Singh, T. Ahmed, and T. Rasheed. Cellular software Future Generation Computer Systems, 29(7), 1645-1660.
defined networking: a framework. IEEE Communications Magazine [111] Alberti, A. M., & Singh, D. (2013, May). Internet of Things:
Communications Standards Supplement, June 2015. Perspectives, Challenges and Opportunities. In International Workshop
[82] H. Wang, S. Chen, H. Xu, M. Ai, and Y. Shi. SoftNet: A Software on Telecommunications (IWT 2013).
Defined Decentralized Mobile Network Architecture toward 5G. IEEE [112] Anderson, T., Peterson, L., Shenker, S., & Turner, J. (2005).
Network, March/April 2015. Overcoming the Internet impasse through virtualization. Computer, (4),
[83] Y. Fengyi, A. Haining, M. Chengli, Z. Jianmin, W. Min. A Flexible 34-41.
Three Clouds 5G Mobile Network Architecture Based on NFV & SDN. [113] De Turck, F.; Boutaba, R.; Chemouil, P.; Bi, J.; Westphal, C., "Guest
China Communications, Supplement No.1 2015. Editors' Introduction: Special Issue on Efficient Management of
[84] M. Sama, L. Contreras, J. Kaippallimalil, I. Akiyoshi, H. Qian, and H. SDN/NFV-Based ZHI14ZHI14SystemsPart I," Network and Service
Ni. Software-Defined Control of the Virtualized Mobile Packet Core. Management, IEEE Transactions on , vol.12, no.1, pp.1,3, March 2015
IEEE Communications Magazine, February 2015. [114] Valdivieso Caraguay, ngel Leonardo, et al. "SDN: Evolution and
[85] M. Chen Y. Qian S. Mao W. Tang X. Yang. Software-Defined Opportunities in the Development IoT Applications." International
Mobile Networks Security. Springer, Science+Business Media New Journal of Distributed Sensor Networks 2014 (2014).
York 2016. [115] Tadinada, V.R., "Software Defined Networks: Redefining the Future of
[86] Ericsson, Telefonica and Ericsson Partner to Virtualize Networks, Internet in IoT and Cloud Era," Future Internet of Things and Cloud
http://www.ericsson.com/news/1763979, PRESS RELEASE (FiCloud), 2014 International Conference on , vol., no., pp.296,301, 27-
FEBRUARY 2014. 29 Aug. 2014.
[87] Alcatel-Lucent, Why Service Providers Need an NFV Platform, [116] Ojo, M., Adami, D., Giordano, S. A SDN-IoT Architecture with NFV
http://www.tmcnet.com/tmc/whitepapers/documents/whitepapers/2014/9 Implementation. 2016 IEEE Globecom Workshops, Dec. 2016.
598-why-service-providers-need-nfv-platform.pdf, 2013. [117] Heinzelman, Wendi Beth. Application-specific protocol architectures
[88] Intel, Intel Data Plane Development Kit (Intel DPDK) Overview Packet for wireless networks. Diss. Massachusetts Institute of Technology,
Processing on Intel Architecture, http://www.intel.com/content/dam/ 2000.
[89] N. Zhang, H. Hmminen. Cost Efficiency of SDN in LTE-based [118] Tie Luo; Hwee-Pink Tan; Quek, T.Q.S., "Sensor OpenFlow: Enabling
Mobile Networks: Case Finland. Department of Communications and Software-Defined Wireless Sensor Networks," Communications Letters,
Networking, Aalto University, 2015. IEEE , vol.16, no.11, pp.1896,1899, November 2012
[90] Alcatel-Lucent virtualized EPC: delivering on the promise of NFV and [119] T. Luo, H.-P. Tan, and T. Q. S. Quek, A compiled list of
SDN. Application note. Alcatel-Lucent, 2014. transportprotocols for wireless sensor networks, May 2012. Available:
[91] Virtualizing the radio access network (RAN): Re-architecting mobile http://www1.i2r.a-star.edu.sg/luot/wsn-tp.pdf
radio access networks for today and tomorrow. Strategic white paper. [120] Zeng, D.; Miyazaki, T.; Song Guo; Tsukahara, T.; Kitamichi, J.;
Alcatel-Lucent, 2015. Hayashi, T., "Evolution of Software-Defined Sensor Networks," Mobile
[92] Aricent: Cloud RAN. Aricent, 2016. Ad-hoc and Sensor Networks (MSN), 2013 IEEE Ninth International
[93] C-RAN: The Road Towards Green RAN. White Paper. China Mobile Conference on , vol., no., pp.410,413, 11-13 Dec. 2013
Research Institute. Diciembre, 2014. [121] Hai Huang; Jiping Zhu; Lei Zhang, "An SDN_based management
[94] Cisco Virtualized Packet Core. http://www. framework for IoT devices," Irish Signals & Systems Conference 2014
cisco.com/c/en/us/solutions/service-provider/virtualized-packet- and 2014 China-Ireland International Conference on Information and
core/index.html. Cisco, 2016. Communications Technologies (ISSC 2014/CIICT 2014). 25th IET ,
[95] Cisco Extensible Network Controller: An OpenDaylight-Based vol., no., pp.175,179, 26-27 June 2013
Controller. Datasheet. Cisco, 2014. [122] Zhijing Qin; Denker, G.; Giannelli, C.; Bellavista, P.;
[96] D. Wenshuan. SoftCom, unified ICT infrastructure enabling Venkatasubramanian, N., "A Software Defined Networking architecture
SDN&NFV. Huawei Technologies Co. Ltd, 2014. for the Internet-of-Things," Network Operations and Management
[97] Enabling Services in the RAN: Complete Pre-Integrated Mobile Edge Symposium (NOMS), 2014 IEEE , vol., no., pp.1,9, 5-9 May 2014
Computing Solution by Juniper Networks and Saguna. Juniper [123] Wu, D., Arkhipov, D. I., Asmare, E., Qin, Z., & McCann, J. A.
Networks. September, 2015. "UbiFlow: Mobility Management in Urban-scale Software Defined
[98] Juniper Networks-Affirmed Networks, Joint Virtualized Mobile Services IoT." INFOCOM, 2015 Proceedings IEEE, May 2015, to appear.
Solution. Solution Brief. Juniper Networks. February, 2016. [124] Patouni, E., Merentitis, A., Panagiotopoulos, P., Glentis, A., &
[99] The what and why of NFV orchestration. Strategic white paper. Nokia, Alonistioti, N. (2013, November). Network Virtualisation Trends:
2016. virtually anything is possible by connecting the unconnected. In Future
[100] Nokia Networks, Aalto University and partners combine NFV and SDN Networks and Services (SDN4FNS), 2013 IEEE SDN for (pp. 1-7).
for virtualized mobile network. Press release. Nokia Aalto University, [125] Chowdhury, NM Mosharaf Kabir, and Raouf Boutaba. "A survey of
2015. network virtualization." Computer Networks 54.5 (2010): 862-876.
[101] M. Paolini. The emergence of the NFV ecosystem: Laying the [126] Omnes, N., Bouillon, M., Fromentoux, G., & Grand, O. (2015,
foundation for a new way to run mobile networks (Report). Senza Fili February). A programmable and virtualized network & IT infrastructure
Consulting, 2014. for the internet of things: How can NFV & SDN help for facing the
[102] Enabling More Versatility and Higher Performance for VNFs with the upcoming challenges. In Intelligence in Next Generation Networks
MaxCore Platform. Artesyn Embedded Technologies. October, 2015. (ICIN), 2015 18th International Conference on (pp. 64-69). IEEE
[103] Cloud RAN Doing More with Less. Solution brief. Artesyn CO. [127] Wood, T.; Ramakrishnan, K.K.; Hwang, J.; Liu, G.; Zhang, W.,
February, 2016. "Toward a software-based network: integrating software defined
[104] Altiostar Solutions. www.altiostar.com. networking and network function virtualization," Network, IEEE,
[105] NECs vEPC Solutions, 2015. vol.29, no.3, pp.36,41, May-June 2015
[106] E. Miller. Brocade helps mobile operator get to 5G with new IP. The [128] Ding, W.; Qi, W.; Wang, J.; Chen, B., "OpenSCaaS: an open service
NEW IP, Febrero 2016. chain as a service platform toward the integration of SDN and NFV,"
http://www.thenewip.net/document.asp?doc_id=721101&site=thenewip Network, IEEE, vol.29, no.3, pp.30,35, May-June 2015
[107] J. Duffy. SDN market could hit $18B by 2018. Network World. [129] Neves, P., Computer Standards & Interfaces (2017),
September, 2014. http://dx.doi.org/10.1016/j.csi.2016.12.008
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 17

[130] SELFNET Framework for Self-Organized Network Management in Electronics and Telecommunications Engineering Department.
Virtualized and Software Defined Networks. https://selfnet-5g.eu/ His main research interests include quality of service, software
defined networking, network virtualization, and resource
allocation.
Alexander Galvis (M2009) is Electronics and
Telecommunications engineer from Cauca University, Lus F. Fletscher is MSc. in Engineering from Universidad
Colombia, and candidate to MSc. in Engineering from Pontificia Bolivariana, Colombia, and MSc. in Information
Universidad Pontificia Bolivariana, Colombia. He is an Technologies and Advanced Telematics from University of
Adjunct Professor with the Electronics and Murcia, Spain. He is an Assistant Professor with the
Telecommunications Engineering Department, University of Electronics and Telecommunications Engineering Department,
Antioquia, Colombia. His main research interests include
wireless and mobile communications, software defined University of Antioquia, Colombia. In 2014, he joined GITA
networking, and resource allocation. (a research group on applied telecommunications) with the
Electronics and Telecommunications Engineering Department.
Juan F. Botero received the Ph.D. degree in telematics He is currently pursuing doctoral studies in engineering at the
engineering from the Technical University of Catalonia, National University, Colombia, and his main research interests
Spain, in 2013. He is an Assistant Professor with the include resource allocation, energy efficiency, software
Electronics and Telecommunications Engineering Department, defined networking, and mobile networks.
University of Antioquia, Colombia. In 2013, he joined GITA
(a research group on applied telecommunications) with the

Vous aimerez peut-être aussi