Making Software-defined Networks Semantic

T. Cooklev

Wireless Technology Center, Purdue University Fort Wayne, Fort Wayne, IN, U.S.A.

Keywords: Ontology, Cognitive Radio, Reconfigurability, Software-Defined Networking, Protocols.

Abstract: This position paper identifies the increasing role of the controller in radios and radio networks. The paper

defines wireless software-defined networks. A new cognitive radio ontology is proposed that is a hierarchical

abstract description of the communication/networking scenario, RF devices, policies, and tasks. Radio

protocols can also be described in a similar way. The ontology provides awareness and supports reasoning by

the controller and applies to any RF device. Directions for future work are also briefly discussed.

1 INTRODUCTION

There are currently several important trends in

wireless radios and networks. We believe that all of

these trends appear as a result of the increasing role of

the controller. Every radio has a controller, which is

responsible for providing and managing the sets of

user interfaces that are necessary to set up and take

down communications sessions. Some of the first

people to think about the expanding role of the

controller were in the Software-Defined Radio (SDR)

community. In a SDR, the controller has to support a

new set of functions that are associated with changing

radio protocols. The original concept of the controller

assumed that a particular fixed radio protocol was to

be “switched in,” therefore the controller was referred

to as a “switcher”.

Cognitive radio has emerged as a concept in the

last ten years. The ITU defines cognitive radio as a

radio that can “obtain knowledge of its operational

and geographical environment, established policies

and its internal state; to dynamically and

autonomously adjust its operational parameters and

protocols according to its obtained knowledge in

order to achieve predefined objectives; and to learn

from the results obtained” (ITU, 2009). Some

cognitive radios address only dynamic spectrum

access and are implemented at layer 2, without a

change in the legacy MAC. These are narrowly-

defined cognitive radios. A cognitive radio must have

domain knowledge of radio communication. Based

on this knowledge, the cognitive engine (CE) can

“dynamically and autonomously” optimize the

various parameters and protocols.

Recently, the area of Software-Defined Networks

(SDNs) has emerged (Bahl et al, 2006; M. Mendonca

et al, 2013). SDN are based on the idea of decoupling

of network control from the forwading plane, i.e. SDN

take the increasing role of the controller to the

networking level. SDN enable network virtualization,

which allows the physical network to be viewed as an

abstract pool of resources. The SDN controller can

handle the configuration of network components

taking into account various policies and applications.

The main motivation for SDNs so far has been flow-

based routing, where flows from different sources are

routed differently. This requires the forwarding

hardware module to be programmable via an open

interface.

It is of considerable interest to extend the SDN

concept to radio networks. Somewhat following ITU’s

terminology, we define wireless SDN as “networks,

where all protocols can be set or altered by software,

excluding changes which occur during the normal pre-

installed and predetermined operation of a network

according to a system specification or standard”. We

recognize that SDNs require in general all protocols to

become software-defined, including physical layer,

MAC, etc. It can said that a SDN is a system-of-

systems, where previously independent components

form a new system with new capabilities. Cognitive

radio networking (CRN) is an example of a new

capability. These CRNs can acquire knowledge about

the network and dynamically and autonomously

adjust their parameters taking into account this

knowledge, pre-defined objectives, and previous

experience. For example, CRNs can respond to

interference, device density, and end-user application

requirements.

Cooklev T..

Making Software-deﬁned Networks Semantic.

DOI: 10.5220/0005558700480052

In Proceedings of the 12th International Conference on Wireless Information Networks and Systems (WINSYS-2015), pages 48-52

ISBN: 978-989-758-119-9

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

Current wireless networks are far from this vision.

Currently, heterogeneous wireless networks are not

software-defined; once configured, it is not possible

to incorporate a new and different RF device without

new hardware and/or new software being installed.

Current networks lack abilities to self-configure. Self-

configuration is desirable at the device and at the

networking level.

After the initial self-configuration, the RF device

shall have sufficient capability to communicate with

other devices and can obtain additional configuration

parameters. The next step after self-configuration is

self-optimization, where the network and its RF

devices can automatically take actions based on the

available information, prior knowledge, policies, and

objectives that have been specified. For example,

when a node drops off the network, traffic is re-routed

around the missing node as necessary to complete the

transmission path. In general, it is desirable to adjust

the parameters of the MAC sublayer of the data link

layer and the physical layer, and all protocols to

achieve a certain objective. One objective can be to

minimize interference. Yet another objective may be

to configure a radio as a relay and in this way extend

the coverage area of a network.

At present, all radios contain an internal

repository of useful information that is accessible

using Simple Network Management Protocol

(SNMP) through the radio’s internal IP address for

devices that are connected to it. This repository is the

radio’s Management Information Base (MIB). The

MIB typically contains information that describes the

frequency, bandwidth, quality of service, interference

or collisions with nearby networks, and so on. This

information is heavily dependent on the physical

layer and the MAC layer of the data link layer of the

given wireless systems. The information available

through the radio’s MIB cannot be understood by

other wireless systems. For example, in a network of

heterogeneous RF devices there will be multiple

MIBs and it is impossible for one RF device to

interpret the MIB values of a different device. The

presence of MIBs do not make devices and networks

software-defined.

Cognitive radio networks require considerable

interaction among the RF devices and the applications

that run on them. The different RF devices must

communicate to the network their observations and

operational states. This information is much richer

than common link status information. For example,

one radio might send a list of all emitters it has

recently sensed to other devices in the network. The

entry for each emitter might include a frequency

range, time, spatial location, and signal format (e.g.,

spread spectrum or narrow-band FM). This requires an

appropriate abstraction, or language. The network also

must communicate its changing operational settings

with all wireless devices.

It is recognized that one of the main bottlenecks

in achieving this vision is the lack of an appropriate

language (Kokar et al, 2008; Cooklev and Cummings,

2008). This language has variously been called a

meta-language, a policy language, a functional

description language, and a network description

language, among others (Kokar et al, 2008; Cooklev

and Cummings, 2008). This language must allow

different types of radios and networks to

autonomously negotiate with each other to specify

and configure themselves in an optimal fashion given

their capabilities, environment, and the objectives of

their users.

A cognitive radio ontology has been developed by

the SDR Forum (Wireless Innovation Forum, 2010).

However, this ontology cannot describe the topology

of a radio. It also tries to define fundamental wireless

communication parameters such as “bit”, “symbol”,

“chipping sequence”, etc., which is at an inappropriate

level of abstraction for describing SDNs. The

ontology of the SDR Forum is mostly used for

adaptive modulation to minimize the size of the bit

error rate (BER). We believe this functionality is best

left to the physical layer. As a result, this ontology is

at the same time not sufficient and adds too much

overhead. Furthermore, parameters such as “symbol”

have different meaning for different radio protocols.

For example, for multicarrier modulation systems

“symbol” has a different meaning than for single-

carrier systems. One approach is to extend the

cognitive radio ontology by providing all possible

symbol definitions. However, it is more important to

address first the question what are the parameters that

should be described by a RF ontology. This question

has not been adequately addressed by the ontology

1.0. We propose a cognitive radio ontology 2.0 that

takes a holistic approach. The operational benefits of

our ontology include seamless interoperability of

heterogeneous RF devices, reduced interference, and

abstraction of device interfaces, which facilitates

assigning tasks to legacy radio devices.

2 ONTOLOGIES

An ontology is a data model that represents a domain,

in our case a wireless networking environment, and is

used to reason about the individuals in the domain and

the relations between them, thus providing a way to

represent knowledge in a standard way. Note that the

MakingSoftware-definedNetworksSemantic

unified modelling language (UML), a language for

specifying software systems, and ontologies

converge.

The Resource Description Framework (RDF) is a

simple ontology language that describes things using

triplets, e.g., subject, predicate, and object (Cooklev

and Cummings, 2008).

An ontology language, such as the Web Ontology

Language (OWL), can be used to described a RF

Device (moving or stationary), a radio transmission

policy, and a task, such as spectrum sensing,

frequency jamming, and so on. An ontology, once

represented in OWL, defines vocabularies for

representing meaning of a subset of domain-

dependent terms and the relationships between these

terms. Using an ontology, information can be

annotated, shared, and reasoned over across

heterogeneous domains, applications, and platforms.

Specifically, the ontology can be used to describe

classes, properties, individuals, and data values. The

language allows us to define relationships between

classes, such as containment. It also allows us to

identify individuals that belong to classes and set their

data and object properties. While the domain of a data

property is a primitive type, such as integer or string,

the domain of an object property is an object. Note

that it is possible for an object to have zero or more

values for a given property and these values do not

need to be of the same type.

3 SEMANTIC RADIO

NETWORKS

We propose a hierarchical description, describing the

communication/networking scenario, RF devices,

policies, and tasks with the following main

parameters.

3.1 Communications/Networking

Scenario

• Setting/terrain

• RF environment

• Interference

• Mobility

• RF device types

• Information type

• Security

• Network topology/NetworkProfile/NeighborList

• QoS parameters

3.2 RF Device

• Time-Of-Day

• Remaining Battery Level / Power spent while

inactive (but powered on)

• Location

• RF front-end parameters

• Digital hardware parameters

3.3 Policies

• regulatory policy

• service provider policy

• user policy

• mission policy

• security policy

• vendor policy, etc.

• spectrum usage policy (spectrum etiquette)

3.4 Tasks

• transmit

• receive

• spectrum sensing

With this abstraction the developed ontology can

describe any signal impinging on the receiver’s

antenna and leverages the VITA 49 standard

(Cooklev and Nishihara, 2013).

Transmitting and receiving can be considered as

tasks. The waveform to use (e.g., GSM or WiFi) is a

parameter of a task. The duration of the task, the start

time, and the frequency range are all other tasks

parameters are recorded in the ontology. The task to

function as a relay can be considered as an ordered

sequence of the transmit and receive tasks. The

topology of wireless networks changes dynamically.

Therefore, it is important to enable self-configuration.

When the topology of the network changes, some

radios may be given the task to begin functioning as

relays.

An important practical consideration is latency. In

general, it can be assumed that the latency is on the

order of tens of milliseconds. Parameters that change

more often are left out of the description.

The ontology provides knowledge, i.e. it makes

the logically centralized controller in a SDN aware of

all parameters. The next step is reasoning based on

the ontology. A reasoning problem is deciding if an

OWL description is consistent and deciding if one

description is subsumed by another. An OWL

reasoner can help us determine if the description

contains any contradicting information. Similarly, an

OWL reasoner can help us determine which

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems

capabilities of a SDR conflict with existing over-the-

air policies.

Modern radio protocols are very complex, but are

not optimal in all scenarios. A physical layer can be

optimized to operate over long range, or high

mobility, or power efficiency, or some combination

of these parameters, to work in different

environments like urban, rural, and so on. Moreover,

the use of different antenna types (such as directional

antennas) may affect the operation of the radio

protocols. Typically, standards groups translate

scenario requirements into technical standards that

work well on average. Fixed physical layers have

options that turn on and off certain features. The

developed ontology takes this process further and

enables all protocols to become software-defined.

Service providers can advertise in a service

registry the descriptions of their capabilities. Every

service has a service profile – what inputs does it

require and what outputs does it provide, and a service

model – how does it work. The ontology must provide

declarative APIs for the automatic execution of the

services. Clients can search using an ontology query

language, interpret these descriptions, and select

appropriate services. This enables dynamic discovery

of services. Automatic service discovery is the

automatic discovery of devices that provide particular

services, without prior negotiations between clients

and service providers. Queries can be made and

answered using appropriate unicast or broadcast

messages. Information messages can be sent

automatically without requests. Their transmission

may be periodical or triggered by certain events.

Heterogeneous nodes can use the ontology to

discover networks and networking opportunities (for

peer-to-peer communication) for user data

transmissions. Without a coexistence mechanism the

nodes searching for networks or networking

opportunities would use technology specific network

search mechanisms, such as scanning the whole band

separately with each technology that the node is

capable and willing to use for user data

communication.

Note that not all devices in the network are

software-defined and/or cognitive (using dynamic

spectrum access). The ontology enables the logically

centralized network controller to be made aware of

legacy devices that cannot communicate using

ontology descriptions. In this way the network

controller can have a global view of the network,

taking into account all RF devices.

4 CONSIDERATIONS FOR

CELLULAR NETWORKS

The ontology in the previous Section is general,

applying to any physical hardware component or

system that interfaces with the RF spectrum. In

special cases such as cellular and WLAN, a number

of parameters are known and therefore do not need to

be described. Cellular networks at present have

significant challenges such as dense deployment,

limited spectrum, consumer demands for data rate,

interworking with WLAN, etc. To address these

challenges current cellular networks, before

establishing a certain capability, employ many

protocol exchanges between a mobile device and

different components of the cellular network such as

base station, radio network control, access gateway,

etc. At best, it is unclear that these exchanges are

optimal. We propose a methodology according to

which cellular networks can better address the

challenges that they are facing. For example, public

land mobile networks (PLMNs) can be provided with

a control plane using such abstract descriptions.

Further investigation of the controller structure and

protocol exchanges is an appropriate topic for further

research.

5 CONCLUSIONS

In this paper we survey the evolution roadmap of

wireless radios and networks. The increasing role of

the controller is identified as the main theme for this

evolution. We advance a comprehensive ontology for

SDN. The ontology describes the network, the RF

devices, their components and protocols that they

support, the policies, and the tasks to be performed.

Note that services are moving from human-to-human

and machine-to-human to machine-to-machine

interactions. Services are available at the “enterprise

level”. A given service, such as voice, can involve

many heterogeneous devices.

The operational benefits of the proposed

technology include:

 Seamless introduction of new RF devices,

 Reduced interference

 Exact description of the RF signal impinging on

the antenna in order to provide complete RF

situational awareness

 Abstraction of device interfaces, which

facilitates assigning tasks to devices.

The implementation benefits of the proposed

ontology include modularity and common interfaces.

MakingSoftware-definedNetworksSemantic

Note that these ontology descriptions may reuse some

of the higher layer functionality – for instance, using

TCP to communicate to a peer process. We do not

consider this a layer violation since the layering of

functionality only applies to data packets. We assume

that these ontology descriptions are sent over a logical

control channel. It can be mapped to a physical

channel in a variety of ways; however this is not

discussed in the paper. It must be noted that in

dynamic spectrum access schemes certain ontology

parameters (such as spectrum occupancy information)

must be delivered before they become outdated. This

problem is related to the way the logical control

channel is mapped to a physical channel and is not

addressed in the paper. We consider the overhead

introduced by the ontology to be small and negligible

compared with high data-rate wireless protocols such

as IEEE 802.11n, LTE-Advanced, etc.

Several topics are identified for further research.

The hierarchical approach can be continued to make

the radio protocols software-defined. Typically,

standards groups translate scenario requirements into

technical standards. We allow in principle this process

to be done automatically. In other words, now there is

a collection of resources (for example, modulation and

coding schemes) from which a physical layer can be

designed. The benefits of the proposed solution are

simpler and faster integration of products from

multiple sources and lower cost of upgrades. This has

been investigated recently for important special cases

such as local area networks and wide-area networks

(Tinnirello et al, 2012; Gallo et al, 2013; De Mil et al,

2014; De Poorter, 2008). Other problems for future

work are the automatic generation of the ontology and

reasoning with inconsistent ontologies.

REFERENCES

Report ITU-R SM.2152, 2009. Definitions of Software

Defined Radio (SDR) and Cognitive Radio System

(CRS).

Mendonca, M., Astuto, B., Nunesy, A., Obraczka, K.,

Turletti, T., 2013. Software Defined Networking for

Heterogeneous Networks, IEEE MMTC E-Letters 8, 3

36-39.

Bahl, P., Chandra, R., Padhye, J., Ravindranath, L., Singh,

M., Wolman, A., Zill, B., 2006. Enhancing the security

of corporate Wi-Fi networks using DAIR. In

Proceedings of the 4th international conference on

Mobile systems, applications and services, MobiSys

’06, ACM.

Kokar, K., Hillman, D., Li, S., Fette, B., Marshall, P.,

Cummings, M., Martin T., Strassner, J., 2008. Towards

a Unified Policy Language for Future Communication

Networks: A Process. In DySPAN’08, Chicago, IL.

Cooklev, T., Cummings, M., 2008. Networking Description

Language for Ubiquitous Cognitive Networking. In

Software Defined Radio Technical Conference,

Washington, DC, USA.

OWL 2 Web Ontology Language Profiles (Second Edition)

http://www.w3.org/TR/owl2-profiles/

Cooklev, T., Nishihara, A., 2013. An Open RF-Digital

Interface for Software-Defined Radios. IEEE Micro.

Wireless Innovation Forum, 2010. Description of the

Cognitive Radio Ontology, WINN-10-S-007.

Tinnirello, I.; Bianchi, G.; Gallo, P.; Garlisi, D.; Giuliano,

F.; Gringoli, F., 2012. Wireless MAC processors:

Programming MAC protocols on commodity

Hardware," In Proceedings IEEE INFOCOM, pp.1269-

1277.

Gallo, P.; Garlisi, D.; Giuliano, F.; Gringoli, F.; Tinnirello,

I.; Bianchi, G., 2013. Wireless MAC Processor

Networking: A Control Architecture for Expressing and

Implementing High-Level Adaptation Policies in

WLANs. IEEE Vehicular Technology Magazine, vol.8,

no.4, pp.81-89.

De Mil, P., Jooris, B., Tytgat, L., Hoebeke, J., Moerman, I.,

Demeester, P., 2014. snapMac: A generic MAC/PHY

architecture enabling flexible MAC design. Ad Hoc

Networks, vol. 17, pp 37-59.

De Poorter, E., Latré, B., Moerman, I., Demeester, P., 2008.

Symbiotic Networks: Towards a New Level of

Cooperation Between Wireless Networks. Wireless

Personal Communications, vol. 45, issue 8, pp 479-

495.

WINSYS2015-InternationalConferenceonWirelessInformationNetworksandSystems