SDN/NFV Architecture for IoT Networks

Angel Leonardo Valdivieso Caraguay

, Patricia Jeanneth Lude

na-Gonz

alez

Rommel Vicente Torres Tandazo

and Lorena Isabel Barona L

opez

Department of Electronics and Computer Science (DCCE), Universidad T

ecnica Particular de Loja, Loja, Ecuador

Department of Informatics and Computer Science (DICC), Escuela Polit

ecnica Nacional, Quito, Ecuador

Keywords:

IoT, NFV, SDN, QoS.

Abstract:

Since the appearance of the Internet of Things concept the research community has been focused on how to

allow an efﬁcient communication and data analysis of millions of devices connected to Internet. However,

the current efforts have not enough in order to cover the IoT requirements and challenges. In recent years,

technologies like SDN, NFV, edge computing, among others, have been pushing not only traditional networks

but also the IoT environments. One of the main concerns is to ensure the quality of the services provided by

IoT devices, for its purpose the decouple of data and control plane, the virtualization of network functions,

advanced data analysis capacity and the ability to share and put closer the services are considered promising

characteristics for this kind of environment.This paper presents an architecture that integrates SDN and NFV

focused on IoT environments and a proof of concept to enhance the quality of services. The experiment

takes advantage of the controller capabilities in order to modify QoE/QoS ﬂows in real time by means the

conﬁguration of SDN-App.

1 INTRODUCTION

The growth of devices connected to Internet has

brought uncertainty and concerns about the real capa-

city of current networks. According the Leading IoT

Gartner report (Gartner, 2017), the number of con-

nected devices will exceed the 20 billions by 2020

and thus the industry and the research community

are introducing novel technologies in order to ensure

the correct and efﬁciently operation of the network.

Software Deﬁned Network (SDN), Network Function

Virtualization (NFV), Self-organized network (SON),

Mobile Edge Network (MEC), cloud and fog compu-

ting, artiﬁcial intelligent, among others, are conside-

red key enabled technologies in this totally networked

world.

The Internet of Things (IoT) could take advantage

from the combination of these technologies in order to

provide a context aware environment (Palattella et al.,

2016),(Li et al., 2015). The main current IoT challen-

ges are related to technology convergence, processing

of huge volume data, network programmability, limi-

ted capacity, energy scarcity, context modeling, rea-

soning, ensuring Quality of Services (QoS), security,

trust and privacy, among others (Sheng et al., 2013),

(Perera et al., 2014). The key idea behind NFV is

to share the network infrastructure between different

physical infrastructures/vendors in order to deploy a

traditional network function as a virtual instance. For

its part, SDN separates the data plane (forwarding

tasks) and control plane (decision and control tasks)

of current network devices, allowing the network pro-

grammability. In the IoT context, the synergy bet-

ween SDN and NFV aims to enhance the control and

deployment of IoT devices or sensors in an efﬁcient

and cost-effective way (Bizanis and Kuipers, 2016),

(Barona L

opez et al., 2015).

Furthermore, IoT devices not only have a critical

problem in terms of network, computing and storage

capacity but also scalability, data access and complex

analysis requirements (D

ıaz et al., 2016). So, IoT

could take advantage from edge and cloud compu-

ting due their facility to provide unlimited resources

(storage, computing and network) closer to the requi-

red service or where the user is located. These resour-

ces can be quickly deployed with a minimal effort or

interaction of the service administrator, establishing

on demand business model.

IoT environments require the processing, corre-

lation and analysis of raw data acquired by differed

devices such as sensors, but because of their limited

resources, these processes must be done outside. In

Caraguay, Á., Ludeña-González, P., Tandazo, R. and López, L.

SDN/NFV Architecture for IoT Networks.

DOI: 10.5220/0007234804250429

In Proceedings of the 14th International Conference on Web Information Systems and Technologies (WEBIST 2018), pages 425-429

ISBN: 978-989-758-324-7

425

this context, advanced analysis capabilities and arti-

ﬁcial intelligent can aid to build knowledge based on

the massive data sending by IoT devices. This ap-

proach lets to offer new kind of applications knowing

as Cloud of Things or Everything as a Service (Ca-

valcante et al., 2016). One of the main issues in this

regard is to ensure the quality of the provisioned servi-

ces. From the Quality of Service point of view, SDN

and NFV play an important role in order to ensure the

service level agreements (SLAs) through their elasti-

city to manage and deploying new network functions

or sensors when the QoS levels are decreasing or if

the user needs additional services. This lets not only

enhance the QoS of the provisioned service but also to

provide a better Quality of Experience (QoE) to end

users. Al the same time, the service and telecommu-

nication providers can be beneﬁted by the reduction

of both capital and operational expenditures (CAPEX

and OPEX).

This paper presents an IoT architecture based on

the combination of SDN and NFV with the objective

to realize proofs of concept related to QoS challenges

in this kind of environments. This document is or-

ganized into four sections, being the ﬁrst of them the

present introduction. Section 2 describes the propo-

sed SDN/NFV Architecture. Section 3 shows the re-

sults of the ﬂow modiﬁcation and its inﬂuence in the

QoS/QoE levels, in real time. Finally the conclusions

are presented in Section 4.

2 SDN/NFV ARCHITECTURE

The global description of the architecture is described

in Figure 1. The proposed architecture follows and in-

tegrates the design principles promoted by ONF (Ong,

2017) and ETSI (Matias et al., 2015). The model is

composed by well-deﬁned four layers: Infrastructure,

Control and Virtualization, Application and Orches-

tration and Management.

The different layers are composed by sublayers

(Figure 2), as follows:

1. Infrastructure Layer. It includes the hardware and

basic software components needed to forwarding

the trafﬁc. In contrast to traditional network de-

vices, IoT devices also originates their own trafﬁc

provided by the internal sensors. In this context, a

dual or hybrid functionality is proposed. IoT no-

des execute traditional forwarding protocols such

as (AODV (Chakeres and Belding-Royer, 2004),

BCHP (Gondaliya and Kathiriya, 2016), DSDV

(Perkins et al., 2001)), among others, depending

of their standard capabilities. If the nodes have

connection with the IoT controller, the nodes ope-

Figure 1: IoT SDN/NFV Architecture.

rate as a data plane waiting for the conﬁguration

provided by control plane. The southbound API

enables the communication between data and con-

trol plane.

2. Control and Virtualization Layer. It provides the

control of the forwarding data behavior and the

virtualized resources for the instantiation of the

NFV Apps. For this purpose, it is composed by

the control plane and virtualized elements. The

control plane has basic control functions. The

Device Manager registers the available nodes that

has connection with the controller and can receive

control plane messages. The unique identiﬁcation

of the nodes (e.g. IP, MAC address) and other

identiﬁcation parameters will be available for the

other modules. Similarly, the Topology Manager

uses the information of the Device Manager and

attempts to map the location and connectivity ca-

pabilities of the devices. The Statistics Manager

listens the communication messages between data

and control plane and infers some basic statistics

information (e.g. bytes sent by node links). Furt-

hermore, the Flow Manager registers the active

data ﬂows in the network.

The administrator also has the capability of in-

clude their own control plane modules for spe-

cial purposes. These applications are known as

SDN Apps. For its part, the virtualized resour-

ces (storage, networking and computing), known

as NFV Apps, are available for their download,

installation or conﬁguration in the virtual infra-

structure. These Apps are considered high level

applications.

3. Application Layer. The different NFV applica-

tions are located on this layer. This layer fol-

lows the virtualization principles applied in com-

puter science, where different user applications

ITSCO 2018 - Special Session on Internet of Things and Smart Communities

426

can be executed in different operating systems

(OS) sharing the same hardware resources. In

this context, the different users can share virtua-

lized resources (storage, network and computing)

to execute their applications in a isolated environ-

ment. Examples of NFV Apps include security

apps, QoE, data analysis, among others.

4. Orchestration and Management. The paradigms

of separate data and control planes, network

function softwarization and resources virtualiza-

tion require the coordination of the different lay-

ers of the infrastructure. For this reason, the Or-

chestration and Management layer operates and

has the possibility of take actions on the other

layers of the infrastructure. It is responsible to

ensure the enough resources for the instantiation

and operation of the VNF Apps and Control Apps.

For this purpose, it is composed by three sublay-

ers: VNF Manager, Orchestrator and Virtual In-

frastructure Manager (VIM). The VIM has a clear

and updated view of the installed infrastructure

and the available virtualized resources (storage,

networking, computing). This information is sent

to the Orchestrator.

For its part, the Orchestrator receives the VIM in-

formation and ensures the good operation and iso-

lation of virtualized elements. Moreover, the Or-

chestrator authorizes the instantiation of Virtual

Network Functions (VNF) only if virtual resour-

ces are available. Then, the VNF Manager orga-

nizes and registers the instantiation of available

VNFs. It also receives the NFV instantiation re-

quests from different customers.

Figure 2: IoT SDN/NFV Architecture.

3 RESULTS

In order to demonstrate the feasibility of the propo-

sed framework, the simulation results are focused on

the behavior of Infrastructure and Control and Vir-

tualization Layer. For this purpose, mininet (Lantz

et al., 2010) emulations were applied. Mininet is the

SDN emulator widely adopted by research commu-

nity. Mininet uses python scripts to emulate custom

network topologies and includes virtual hosts, OF-

enabled switches and virtual links. However, the emu-

lation performance in real time applications is depen-

dent on the underlying host system capacity. In this

context, the use of small scale topologies guarantee

the accuracy results and CPU/memory isolation.

The topology is emulated using a server Acer

Swift 3 (Intel i7, 2.7 Ghz, 8GB) running a virtual ma-

chine with Linux Ubuntu 16.04. The topology (Fi-

gure 3) emulates a common data center with core,

aggregation and edge OF-enabled switches (s1-s7).

The switches are connected to the SDN Floodlight

(Floodlight Controller, ) controller. Floodlight pro-

vides a REST-API to request the main SDN services

required to implement SDN-Apps. The present ex-

periment uses the QoS-App (Wallner and Cannistra,

2013), (Wallnerryan, ). Each edge switch (s4-s7) is

connected with two hosts (h1-h8). The virtual hosts

represent the IoT devices and generate video strea-

ming information. The virtual hosts use a VLC server

to stream the video ﬁle “highway cif ”(Telecommuni-

cation Networks Group, ) using RTP/UDP. The video

is encoded in MPEG4 (highway cif.mp4) with a size

ﬁle of 4.23 MB and a resolution of 352 x 288. The vi-

deo is composed by 2000 frames and a total duration

of 66 seconds.

Figure 3: Test Topology.

The objective of the experiment is to demonstrate

the controller capability to customize the switches be-

havior without affect the normal operation of the net-

work. The experiment uses two streaming ﬂows. The

ﬁrst streaming (w/o-QoS) is sent from h1 to h8 and

the second stream (w-QoS) sents the video from h2 to

h7. Both videos are simultaneously sent through the

network. During the streaming, after 45 seconds, the

QoS-App is downloaded and installed in the control-

SDN/NFV Architecture for IoT Networks

427

ler. Then, the QoS-App is automatically conﬁgured

to control the bandwidth of the links depending of the

ﬂow. The w-QoS streaming (ﬂow from h2 to h7) is

limited with a maximal datarate of 2Mbps and the se-

cond ﬂow w/o-Qos (h1 to h8) is limited to a maximal

datarate of 0.4 Mbps. Then, the received streams are

saved in different video ﬁles.

The received ﬁles are processed using the Evalvid

Tool (Department of Telecommunication Systems, ),

(Klaue et al., 2003). For this purpose, the ﬁles are

decoded as .yuv ﬁles and Evalvid evaluates the Peak

Signal to Noise Ratio (PSNR) and the Structural Si-

milarity Index Metric (SSIM) between both, sent and

received streams. Since several processes are execu-

ted simultaneously in the same VMs (mininet, Flood-

light, VLC, Evalvid), slight variations between test

are depicted depending on the CPU and memory re-

sources. For this reason, the Monte-Carlo method is

used. That is, the scenario is tested 20 times and the

corresponding average is evaluated.

The results of the experiments are depicted in the

Figures 4, 5 and 6. With the purpose of reduce distor-

tions, the trend line with an average of 20 frames is

depicted. Figure 4 shows the PSNR average against

the number of frames. The red line (solid) represents

the w-QoS streaming and the blue line (dotted) re-

presents the w/o-QoS. At the beginning of the expe-

riment, both streaming ﬂows have similar behavior

(best effort). For this reason, the average PSNR is

clearly similar between them. The vertical black line

depicts the point at which the QoS-App is downloa-

ded and conﬁgured in the Floodlight controller. As

expected, after the QoS-App conﬁguration, the be-

havior of the switches are automatically modiﬁed to

identify the ﬂows and assign different QoS policies.

As a result, the h2-h7 trafﬁc shows better levels of

PSNR compared with h1-h8 ﬂow. In this context, the

average w-QoS is 26,54 dB against 23,73 dB of the

w/o-QoS stream.

Figure 4: Peak Signal to Noise Ratio.

The Figure 4 shows a considerable trough in case

plot. The experiments show that the unexpected ef-

fect is mainly caused by the fast moving scenes in this

time. The network load is increased and the PSNR

decreases.

Figure 5: Structural Similarity Index Metric.

For its part, the Figure 5 outlines the SSIM of

the test ﬂows. The average SSIM of w-QoS is 0,814

against the w/o-QoS streaming with 0,738. Finally,

Figure 6 summarizes the above results with the cal-

culation of MOS (Mean Opinion Score). The MOS

(ITU, 1996) attempts to estimate the grade of accepta-

bility of the user (QoE). The scale is proposed as fol-

lows: (5) excellent, (4) good, (3) fair, (2) poor and (1)

bad. Using this background information, the average

MOS during the streaming is calculated. The average

for w/o- QoS is 2,302 against 2,755 of the w-QoS stre-

aming. The obtained results demonstrates the con-

troller capability to dynamically modify the network

behavior and balance the data ﬂows depending of the

user requirements.

Figure 6: Mean Opinion Score (MOS).

4 CONCLUSIONS

The present work analyzes the advantages that SDN

and NFV paradigms introduce in IoT environments.

Moreover, a SDN/NFV architecture for IoT networks

is proposed. In this context, the controller capabi-

lity to dynamically control the network behavior is

ITSCO 2018 - Special Session on Internet of Things and Smart Communities

428

tested. The mininet based test topology and the vi-

deo streaming analysis demonstrate that the ﬂoodlight

controller is capable to modify the QoS/QoE of diffe-

rent ﬂows in real time. Therefore, the challenges for

future research include the balance and orchestration

of virtual resources for IoT environments. Similarly,

the optimization of algorithms for real streaming in

SDN/NFV architectures is challenging.

ACKNOWLEDGMENT

The authors are with the Group of Robust, Sustainable

and Secure Networks (SRSNet), Department of Elec-

tronics and Computer Science (DCCE), Universidad

ecnica Particular de Loja (UTPL), Loja, Ecuador.

REFERENCES

Barona L

opez, L. I., Valdivieso Caraguay,

A. L., Villalba,

L. J. G., and L

opez, D. (2015). Trends on Virtualisa-

tion with Software Deﬁned Networking and Network

Function Virtualisation. IET Networks, 4(5):255–263.

Bizanis, N. and Kuipers, F. A. (2016). SDN and Virtuali-

zation Solutions for the Internet of Things: A Survey.

IEEE Access, 4:5591–5606.

Cavalcante, E., Pereira, J., Alves, M. P., Maia, P., Moura,

R., Batista, T., Delicato, F. C., and Pires, P. F. (2016).

On the Interplay of Internet of Things and Cloud Com-

puting: A Systematic Mapping Study. Computer

Communications, 89:17–33.

Chakeres, I. D. and Belding-Royer, E. M. (2004). AODV

Routing Protocol Implementation Design. In Distri-

buted Computing Systems Workshops, 2004. Procee-

dings. 24th International Conference on, pages 698–

703. IEEE.

Department of Telecommunication Systems. Evalvid. Avai-

lable at http://www.tkn.tu-berlin.de/menue/research/

evalvid.

ıaz, M., Mart

ın, C., and Rubio, B. (2016). State-of-the-

art, Challenges, and Open Issues in the Integration of

Internet of Things and Cloud Computing. Journal of

Network and Computer Applications, 67:99–117.

Floodlight Controller. Open Source Software for

Building Software-Deﬁned Networks. Availa-

ble at http://www.projectﬂoodlight.org/ﬂoodlight

(2018/04/19).

Gartner (2017). Leading IoT Gartner Report.

Gondaliya, N. and Kathiriya, D. (2016). Map Based DTN

Architecture and an Efﬁcient Routing Protocol in De-

lay Tolerant Networks for Post Disaster Situation. In-

ternational Journal of Computer Science and Infor-

mation Security, 14(8):980.

ITU (1996). Methods for Subjective Determination of

Transmission Quality. Recommendation P.800, Inter-

national Telecommunication Union, Geneva.

Klaue, J., Rathke, B., and Wolisz, A. (2003). Evalvid–A

Framework for Video Transmission and Quality Eva-

luation. In International conference on modelling

techniques and tools for computer performance eva-

luation, pages 255–272. Springer.

Lantz, B., Heller, B., and McKeown, N. (2010). A Network

in a Laptop: Rapid Prototyping for Software-deﬁned

Networks. In Proceedings of the 9th ACM SIGCOMM

Workshop on Hot Topics in Networks, page 19. ACM.

Li, S., Da Xu, L., and Zhao, S. (2015). The Internet of

Things: a Survey. Information Systems Frontiers,

17(2):243–259.

Matias, J., Garay, J., Toledo, N., Unzilla, J., and Jacob, E.

(2015). Toward an SDN-enabled NFV Architecture.

IEEE Communications Magazine, 53(4):187–193.

Ong, L. (2017). ONF SDN Architecture and Standards for

Transport Networks: Control Architecture and Net-

work Modeling I. In Optical Fiber Communication

Conference, pages M2H–1. Optical Society of Ame-

rica.

Palattella, M. R., Dohler, M., Grieco, A., Rizzo, G., Tor-

sner, J., Engel, T., and Ladid, L. (2016). Internet of

Things in the 5G era: Enablers, Architecture, and Bu-

siness Models. IEEE Journal on Selected Areas in

Communications, 34(3):510–527.

Perera, C., Zaslavsky, A., Christen, P., and Georgakopoulos,

D. (2014). Context Aware Computing for the Internet

of Things: A Survey. IEEE communications surveys

& tutorials, 16(1):414–454.

Perkins, C. E. et al. (2001). Ad Hoc Networking, volume 1.

Addison-wesley Reading.

Sheng, Z., Yang, S., Yu, Y., Vasilakos, A., Mccann, J., and

Leung, K. (2013). A survey on the IETF Protocol

Suite for the Internet of Things: Standards, Challen-

ges, and Opportunities. IEEE Wireless Communicati-

ons, 20(6):91–98.

Telecommunication Networks Group. Evalvid Video File.

Available at http://www2.tkn.tu-berlin.de/research/

evalvid/cif/highway.

Wallner, R. and Cannistra, R. (2013). An SDN Appro-

ach: Quality of Service using Big Switchs Floodlight

Open-source Controller. Proceedings of the Asia-

Paciﬁc Advanced Network, 35:14–19.

Wallnerryan. Floodlight with QoS module. Available

at https://github.com/wallnerryan/ﬂoodlight-qos-beta

(2018/04/19).

SDN/NFV Architecture for IoT Networks

429