Cloud Description Ontology for Service Discovery and Selection

Molka Rekik

, Khouloud Boukadi

and Hanˆene Ben-Abdallah

1,2

Mir@cl Laboratory, Sfax University, Sfax, Tunisia

King Abdulaziz University, Jeddah, K.S.A.

Keywords:

Ontology, Cloud, Services, Functional Properties, Non-functional Properties, Discovery, Selection.

Abstract:

Cloud computing is a model delivery of infrastructure, platform and software services over the net. The lack

of standardization of heterogeneous Cloud service descriptions makes service discovery and selection very

complex tasks for Cloud users. To ease this complexity, it is crucial to describe the various Cloud service

pertinent information in a homogeneous model. To provide for such model, this paper offers to contributions.

First, it proposes a Cloud description ontology that integrates service descriptions obtained from heteroge-

neous sources. The ontology model accounts for the functional and non-functional properties, attributes and

relations of infrastructure, platform and software services in order to facilitate the discovery and selection of

suitable Cloud services. Second, this paper presents a qualitative and quantitative evaluation of the proposed

ontology. For the qualitative evaluation, we use a reasoner to identify defects in the ontology, whilst for the

quantitative evaluation we compare our search results with those obtained through a web search engine.

1 INTRODUCTION

Cloud computing technology can be viewed as a set

of software services (SaaS), platform services (PaaS)

and infrastructure services (IaaS) offered to users over

the net. It has the advantage of offering users scal-

able and elastic service capabilities, at a reduced cost,

speeding up the deployment of their businesses. In-

deed, in the SaaS model, a user gets the provider ap-

plications that are hosted and run in cloud infrastruc-

ture; in the PaaS model, the user gets the platform that

is running in the cloud infrastructure for creating or

deploying the applications; and in the IaaS model, the

user gets the virtual resources in which an execution

environment can be deployed to run an application.

A Cloud federation is a new paradigm where a

set of Cloud providers are grouped in order to offer

a wider range of resources/services and, hence, en-

large their market shares (Rochwerger et al., 2009),

(Buyya et al., 2010). However, Cloud federations are

hindered by the lack of a standard language for de-

scribing providers’ offers making it difﬁcult for end-

users to discover and select an appropriate provider

within a federation. In fact, service providers de-

scribe similar services with different manners (differ-

ent pricing policies, different quality of services, etc.);

such heterogeneous descriptions complicates service

comparison and selection, and may create interop-

erability problems among multiple providers (espe-

cially when migrating from a provider to another).

To overcome the problems stemming from heteroge-

neous service descriptions within a Cloud federation,

we propose to deﬁne a common ontology for Cloud

service description. This ontology is especially use-

ful when users want to search for domain speciﬁc in-

formation (Ruthven and Lalmas, 2003), (Hoang and

Tjoa, 2006), (Bhogal et al., 2007). It is a means to

share common understanding about the structure of

Cloud service descriptions while covering all Cloud

service layers (IaaS, PaaS and SaaS), and to represent

semantically and structurally Cloud providers’ infor-

mation.

It is worth noting that, there exists some indus-

trial standards such as (NIST, 2013), (Forum, 2010),

(Behrendt et al., 2011) and research projects such

as (Beom et al., 2011), (Grozev and Buyya, 2012),

(Moscato et al., 2011), (Youseff et al., 2008), (Zhang

et al., 2012) that treat the aforementioned problems;

unfortunately, current propositions focus only on the

IaaS layer: enabling uniﬁed access to existing IaaS

services to deal with the heterogeneity of the Clouds.

That is, they do not offer solutions to the PaaS and

the SaaS layers. In fact, our work is the ﬁrst to treat

the three layers together. It relies on existing stan-

dards, Cloud providers’ catalogs and some research

projects to deﬁne an ontology that accounts for a set of

functional and non-functional conﬁguration proper-

ties. The ontology provides for service selection that:

Rekik M., Boukadi K. and Ben-abdallah H..

Cloud Description Ontology for Service Discovery and Selection.

DOI: 10.5220/0005556400260036

In Proceedings of the 10th International Conference on Software Engineering and Applications (ICSOFT-EA-2015), pages 26-36

ISBN: 978-989-758-114-4

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

in the IaaS layer, uses the characteristics of the virtual

machine (VM) (e.g., the CPU type, the memory size,

the operating system, etc.); in the PaaS layer, refers

to the characteristics of the platform (e.g., program-

ming language, libraries, databases, etc.); and in the

SaaS, deals with the application characteristics (e.g.,

usability, scalability, multi-tenancy, etc.).

The rest of the paper is organized as follows. Sec-

tion 2 describes works pertinent to ontology for Cloud

computing. In Section 3, we present our ontology to

describe Cloud services, which we evaluate quantita-

tively and qualitatively in Section 4. Finally, in Sec-

tion 5, we summarize the status of the presented work

and outline its extensions.

2 ONTOLOGY USAGE IN CLOUD

COMPUTING

In order to standardize the representation of Cloud

computing services, various works use an ontology.

In (Youseff et al., 2008), the authors establish the

knowledge domain of the area of Cloud computing

and its relevant components. They discuss each layer

strengths, limitations as well as dependence on other

layers. In addition, they deﬁne a general taxonomy

that captures the inter-relations between the different

Cloud components. This work facilitates the under-

standing of the Cloud environment and its key con-

cepts. The identiﬁed concepts cover the three Cloud

layers and they are deﬁned and generated from stan-

dards. However, the informal description of the con-

cepts in this work does not provide for its direct ap-

plication in a selection method. This motivated sev-

eral propositions to formalize the Cloud concepts in

an ontology that facilitates service discovery.

As Gruber (Gruber, 1995), the ontology provides

an explicit speciﬁcation of a conceptualization. So, an

ontology is a language used for representing formally

the available information of a speciﬁc domain provid-

ing a hierarchy of concepts which can be manipulated

by the users.

In (Parhi et al., 2014), the authors propose

a semantic-based service description and discovery

framework using the multi-agent approach. The

framework mainly assists users to discover suitable

service on-demand based on a Cloud service descrip-

tion ontology. Adopting a similar approach, the au-

thors in (Ghoneim and Tolba, 2014) propose a frame-

work composed of two main components: controller

and cloud ontology blackboard environment. Both

components use an ontology to discover the best

suited set of responses for the consumer requests. In a

similar way, in (Bernstein and Vij, 2010), the authors

deﬁne an ontology based catalogs that describes fea-

tures and capabilities of resources offered by Cloud

providers, such as CPU, storage, security, and com-

pliance capabilities. The main aim of this work is

to provide a federated Cloud environment. Also, the

authors in (Beom et al., 2011) propose an ontology

based resource management for Cloud Computing.

They select the best Cloud resources based on the ob-

jective of the Service Level Agreement (SLA). Over-

all, the aforementioned mentioned ontologies sup-

port only a few number of concepts in the Cloud

IaaS level. In addition, they do not deﬁne the non-

functional elements which are important information

during the selection process to satisfy user require-

ments.

In (Zhang et al., 2012), the authors propose the

CoCoOn ontology for describing Cloud infrastruc-

ture services while detailing the functional and non-

functional properties. Using CoCoOn, they imple-

ment a service discovery system to search for avail-

able infrastructure services. The search method relies

on the typical QoS criteria by adopting a Web service

approach. In addition, because CoCoOn covers only

the IaaS level concepts, the search considers only the

virtual machines as Cloud resources.

An interesting initiative in this area is proposed

by the mOSAIC project (Moscato et al., 2011). The

project analyzes existing standards and proposes an

ontology for semantic retrieval and composition of

Cloud services. It provides a platform and proposes

a set of APIs for solving multiple Clouds interoper-

ability problems. In addition, it treats the functional

and non-functional properties. However, the ontology

proposed in mOSAIC considers the IaaS Cloud layer

and neglects the other Cloud service layers (SaaS and

PaaS).

It is worth mentioning that, in all of the above

mentioned published works, the authors do not show

an evaluation to improve the use of their proposed on-

tologies.

In the herein presented work, we build upon the

above mentioned works to propose a comprehensive

Cloud service description ontology that serves as a ba-

sis for the discovery and selection of Cloud providers

and services in the context of a Cloud federation.

More speciﬁcally, we propose an ontology that covers

both functional and non-functional aspects of Cloud

services, and at the three layers of Cloud (IaaS, PaaS

and SaaS). In addition, we evaluate our proposed

work, ﬁrst, by using a reasoner to examine its logical

or conceptual coherence and, secondly, by comparing

the search results it offers with those of Web search

engine.

CloudDescriptionOntologyforServiceDiscoveryandSelection

3 CLOUD DESCRIPTION

ONTOLOGY

The work presented in this paper is part of an on-

going project where a Cloud service description on-

tology is the building block for the scheduling of

business processes in a Cloud federation environment

(i.e., the task-resource allocation). In our previous

work (Rekik et al., 2014), we presented a context-

based categorization in order to trigger the on-demand

resource provisioning across multiple infrastructures

providers. We deﬁned a task-related context, user-

related context and resource-related context. This

paper presents a Cloud service description ontol-

ogy to model the resource-related context (i.e., the

resource-related context is explicitly deﬁned by Cloud

providers, it is mainly related to the resource/service’s

characteristics. Generally, this information models

the competitive advantage that Cloud providers may

have over one another) in order to facilitate service

(resource) selection from a set of available ones.

To construct the ontology, we proceeded in two

steps. First, we identiﬁed its concepts by analyzing

standards and proposals from the literature. Secondly,

we followed the design principles, presented on (Gru-

ber, 1995), which are objective criteria for guiding

and evaluating ontology designs; they include clar-

ity, minimal encoding bias, extendability, coherence

and minimal ontological commitments. In addition

to these principles, we followed name standardization

which, as suggested in (Cesar et al., 1998), eases the

understanding of the ontology. Following these prin-

ciples, we obtained the concepts shown in Figure 1

and which we detail next. To create the ontology,

we used the prot`eg`e

editor which is the most widely

used, freely available, platform independent technol-

ogy for developing and managing terminologies, on-

tologies and knowledge bases. It is written in Java and

uses Swings to create the complex user interface.

Figure 1: The Top Level Concepts.

3.1 The Actor Class

The Actor class presents the different actors which

http://protege.stanford.edu/

participate in Cloud systems. We distinguish be-

tween:

• According to the mOSAIC (Moscato et al., 2011),

the Developer which develops the Cloud services

and the Administrator which conﬁgures them.

• As NIST (NIST, 2013), the Broker which moni-

tors, negotiates, schedules, etc. these Cloud ser-

vices.

• As IBM (Behrendt et al., 2011), the Consumer

which uses the service and the Provider which

offers/has the Cloud service. The NIST (NIST,

2013) classiﬁes the consumer as IaaS, PaaS and

SaaS

Consumer.

• According to the NIST (NIST, 2013), the

PaaS

Consumer can be an Application Developer

who designs and implements services (application

software), an Application

Admistration who con-

ﬁgures and monitors the performances, an Ap-

plication

Deployer who publishes services into

the Cloud system and an Application

Tester who

runs and tests services. In similar, we classify

three types of providers namely IaaS, PaaS and

SaaS

Provider.

3.2 The Deployment Model Class

Each provider follows/has a deployment model.

By referring to NIST (NIST, 2013), the Deploy-

ment

Model class includes Private, Public, Hybrid,

Community and Federation sub-classes. In private

model, the services are provisioned for exclusive use

by a single organization. In community model, the

services are provisioned for exclusive use by a spe-

ciﬁc community of consumers from organizationsthat

have shared concerns. The services are provisioned

for open use by the general public in the public model.

The hybrid model is a composition of two or more dis-

tinct Cloud models (private, community, or public).

In our work, we haveextended this class by adding

the federation subclass (Figure 2). In the federa-

tion model, the services are provisioned by a group

of Cloud providers that voluntarily collaborate with

each other to exchange resources (Grozev and Buyya,

2012). The federation model has a type and an ar-

chitecture. The federation type can be vertical which

spans multiple levels, or horizontal which takes place

on one level of the cloud stack. The federation ar-

chitecture can be installed with broker (centralized)

where there is a central broker that performs and fa-

cilitates the resource allocation or without broker (de-

centralized) where the cloud providers communicate

and negotiate directly with each other without media-

tors like the peer to peer architecture.

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

Figure 3: The Virtual Machine’s Characteristics.

Figure 2: The Deployment Models.

3.3 The Essential Characteristics Class

Each provider has an essential characteristics deﬁned

by the NIST standard (NIST, 2013). In fact, in our

work, the Essential

Characteristics class includes the

following sub-classes.

• On

Demand Self Service: The provider offer

computing capabilities, such as server, network

and storage as needed automatically without re-

quiring human interaction.

• Broad

Network Access: The resources’ capabil-

ities are available over the network and accessed

through standard mechanisms that promote use by

heterogeneous thin or thick client platforms.

• Resource

Pooling: The resources are pooled to

serve multiple consumers using a multi-tenant

model, with different physical and virtual re-

sources dynamically assigned and reassigned ac-

cording to consumer demand.

• Measured Service: The Cloud systems automati-

cally leverage a metering capability at some level

of abstraction appropriate to the type of service.

• Rapid

Elasticity: The resources’ capabilities can

be elastically provisioned and released to scale

rapidly outward and inward commensurate with

demand.

3.4 The Service Class

The NIST (NIST, 2013) deﬁnes the services offered

by providers as SaaS, PaaS and IaaS. Each provider

offers a service that is in reality a virtualized resource.

That is, an IaaS provider offers either a virtual ma-

chine or a storage space, a PaaS provider offers a plat-

form, whereas a SaaS provider offers a Software or

application.

1. IaaS

Service:

• Virtual

Machine class: By accessing to IaaS

providers’ catalogs (Cloud, 2013), (Henry,

2009), (Russell and Cohn, 2012), we notice that

a VM has a location and a type. Depending

to the type, we distinguish the following VM’s

characteristics a(shown in Figure 3:

– Network

Resource: The bandwidth is a net-

work resource which has a Delay, Through-

put, Protocol and Price

Per Data Transfer.

– Storage Space: The Memory is the storage

space which has Size and Allocation Strategy.

– Computational Resource: The CPU is a com-

putational resource. It has Core(s), Frequency

and Architecture.

– OS: The operating system of a VM.

– Use Case: While examining some Cloud

providers (Cloud, 2013), (Henry, 2009), we

notice that a VM has a use case such as Op-

timized

Memory, Optimized Calculation and

General Use.

– Price Per Hour: The VM’s allocation price

per hour.

– Price Per Instance: The VM’s allocation

price per instance.

Just to note here that, ﬁrst, the relation between

the VM, Location and Type classes must be a

part of relation. Indeed, wherever Type and Lo-

cation exist, it is as part of VM, and their pres-

CloudDescriptionOntologyforServiceDiscoveryandSelection

ence implies the presence of the VM. Similarly

for Type class and its all sub-classes. To de-

scribe a VM, we categorize functional and non-

functional properties which are deﬁned in Sub-

section 3.5 (Figure 4).

Figure 4: The Virtual Machine Description.

• Storage

Space: A Storage space can be pro-

vided with or without a VM. This class

covers Distributed

File System (is a ﬁle sys-

tem that allows access to ﬁles across the

multiple hosts within the Cloud), Repli-

cated

Relational Database (is a distributed

database which allows the copy of data from a

database in one resource to a database in an-

other) and Disk sub-classes. Like the Type

(of VM) class, these sub-classes are part of

Storage

Space class. The storage space has

a Price

Per Data Size, Storage Format, Stor-

age Size Min (the minimum storage capacity)

and Storage

Size Max (the maximum storage

capacity) (shown in Figure 5).

Figure 5: The Storage Space Class.

2. PaaS

Service: The Platform class includes IDE

(an integrated development environment), API

(an Application Programming Interface), Appli-

cation

Server (an environment where applications

can run), Web

Server (server used to publish

websites) and Database (an organized collection

of data) subclasses. These all cited sub-classes

are part of Platform class. Like the IaaS re-

source, the PaaS resource has a Price

Per Hour

and PaaS Service Description which includes the

functional and the non-functional properties.

These properties will be detailed in Subsection

3.5.

3. SaaS Service: The Software has a

Price Per Hour, Price Per License,

Price

Per Instruction, Price Per User Number

and a SaaS Service Description which includes

a functional and non-functional properties which

are detailed in Subsection 3.5.

3.5 The Property Class

The Property subclasses contain all elements needed

for describing characteristics of Cloud resources.

We distinguish the Functional

Property and the

Non

Functional Property subclasses. The ﬁrst de-

ﬁnes the service function (i.e., what a service is sup-

posed to accomplish) while the second speciﬁes the

service quality or criteria. For each service provided

(IaaS, PaaS and SaaS), we deﬁne a set of functional

and non-functional properties.

3.5.1 The Functional Property Class

1. The Functional

IaaS Property class: this class in-

cludes the Functional VM Property and the Func-

tional

StorageSpace Property sub-classes.

(a) The Functional VM Property class: in this

class, we classify the following subclasses

(shown in Figure 6):

• The Virtualization class: it is a mecha-

nism for operating multiple systems, servers,

or applications on the same physical server

as a virtual machine. This class covers

the Full-Virtualization (virtualization of the

OS without any modiﬁcation) and the Para-

Virtualization (virtualization of the OS with

modiﬁcation).

• VM

Management class: it covers Backup,

Recovery and Scheduling. The ﬁrst consists

on creating data images to storage while the

second consists in recovering data in disasters

or failures. The scheduling is the allocation of

cloud resources to consumers.

• VM

Monitoring class: it detects slow or fail-

ing components, such as overloaded or failed

VMs.

• Cloud

Federation class: it allows a set of

Cloud providers to federate together.

• Replication class: it involves sharing informa-

tion to ensure consistency between redundant

resources.

• Migration class: it consists in moving from

the VM overloaded to another underloaded.

(b) The Functional StorageSpace Property class:

such as the Functional VM Property class,

this class covers the Virtualization, StorageS-

pace

Management, StorageSpace Monitoring,

Cloud

Federation, Replication and Migration

sub-classes.

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

Figure 6: The Virtual Machine Functional Property.

Figure 7: The PaaS Functional Property.

2. The Functional

PaaS Property class: its sub-

classes are described in Figure 7. We dis-

tinguish the Deployment, the Development,

the Testing, the Execution, the Replication,

the Cloud

Federation, the Platform Management

which includes the Database and Applica-

tion

Management and the Platform Monitoring

subclass which includes Application

Monitoring

and Database Monitoring.

3. The Functional

SaaS Property class: we deﬁne

the following sub-classes (Figure 8): ERP (is an

Enterprise Resource Planning system which au-

tomates and tracks a variety of business func-

tions across different departments of an organi-

zation), CRM (is a Customer Relationship Man-

agement which provides good customer manage-

ment, opportunities and altogether better cus-

tomer support), Humain Capital Management (it

evolves and automates processes such as payroll,

performance assessments, recruitment and train-

ing), Collaborative

Tools (it allows users to col-

laborate and to work in groups within and be-

tween companies, Content Management (it is a

process enabling company, agency or organiza-

tion to obtain, organize, store and deliver infor-

mation essential to its operation in the most efﬁ-

cient manner), Mailing (it is a mail management

Figure 8: The SaaS Functional Property.

application), Social

Network (a social application

that enables users to interact and share data) and

Video

Gaming (an interactive application that is

used for entertainment, role playing and simula-

tion).

3.5.2 The Non Functional Property Class

Basically, non-functional requirements relate to the

qualities of the service (QoS) that cut across user fac-

ing features, such as security, reliability, availability,

etc. In this work, we present the QoS properties while

referring on the ISO-9126 (ISO/IEC, 2001) and non-

functional properties which are speciﬁc to each ser-

vice (namely IaaS, PaaS and SaaS).

1. The Non

Functional IaaS Property class: this

class includes the Non Functional VM Property

and the Non Functional StorageSpace Property

subclasses.

(a) The Non

Functional VM Property class: in

this class, we classify the following subclasses

(shown in Figure 9):

• VM

Auto Scaling class: In (Alhamad et al.,

2010), the elastic provisioning is deﬁned as a

concept in which computing resources can be

scaled up and down easily by the cloud ser-

vice provider. According to this deﬁnition, we

add the following subclasses: VM

Scale Up

(the maximum number of VMs provided for

one user) and VM Scale Down (the minimum

number of VMs provided for one user).

• Processing

Speed class: the processing speed

of the CPU.

• Data

Transfer Speed class: the processing

speed of the network.

• QoS

IaaS class: it includes the Integrity (the

assurance that the information can be ac-

cessed or modiﬁed only by those authorized to

do), Security, Flexibility (the ability to simply

and instantly manage cloud services), Compli-

ance (the services are conform to standards),

Fault

Tolerance (the ability to manage auto-

CloudDescriptionOntologyforServiceDiscoveryandSelection

Figure 9: The Non-Functional Virtual Machine Properties.

Figure 10: The Non-Functional Storage Space Properties.

mated way component failures) and Availabil-

ity (the services are accessible).

(b) The Non Functional StorageSpace Property

class: in this class, we classify the following

sub-classes (shown in Figure 10):

• Access

Type class: it indicates the access type

of information that can be either read and/or

written.

• Access

Time class: it describes the informa-

tion storage time.

• IOPS class: it includes the input and output

operations per second.

• QoS

IaaS class: it includes the In-

tegrity, Security, Flexibility, Compliance,

Fault

Tolerance and Availability properties.

2. The Non

Functional PaaS Property class:

this class includes Number

Developers, Lan-

guage Of Development, Platform Type, Plat-

form

Architecture and QoS PaaS which includes

the Portability (the usability of the same ap-

plication in different environments), Security,

Flexibility (the ability to adjust in needs), Com-

pliance, Fault

Tolerance (the ability to continue

when a resource failure occurs) and Availability

properties (Figure 11).

3. The Non

Functional SaaS Property class: refer-

ring to (KPMG, 2013), the Multi Tenancy (mul-

tiple users can simultaneously access the same

service), the Integration (integrate the data of

users in the SaaS), the No

Vendor Lock In, the

Easy

Customization (each user can easily cus-

tomize applications to ﬁt their business pro-

Figure 11: The Non-Functional PaaS Properties.

Figure 12: The Non-Functional SaaS Properties.

cesses without affecting the common infrastruc-

ture), the No

Large Up Front Investment (allow-

ing customers to avoid the large initial investment

in IT infrastructure), the Contract

Length (the pe-

riod that the software product is legally available

for a client) and the QoS

SaaS which includes the

Efﬁciency (the ability to accomplish a job with

an optimal performance), Satisfaction, Usability,

Adaptability (the ability to adjust in needs), Secu-

rity, Flexibility, Compliance, Fault

Tolerance and

Availability properties (Figure 12).

3.6 Instances Creation

In this subsection, we associate real world instances to

the proposed concepts by referring to a set of Cloud

providers’ catalogs available on the Internet (such as

Amazon EC2

, Rackspace

, HP Cloud

, etc.). We

present an extract of instances related to the IaaS,

PaaS and SaaS services while focusing on some of

functional and non-functional properties (Tables 1, 2,

3).

4 ONTOLOGY EVALUATION

Once constructed, the ontology needed to be evalu-

aws.amazon.com

www.rackspace.com

www.hpcloud.com

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

Table 1: An Extract of IaaS Individuals.

Domain Object Property Range

AmazonEC2 hasIaaSService VM11, ...

MicrosoftAzure hasIaaSService VM21, ...

... ... ...

VM11 hasType XLarge

hasPricePerHour 0.64$

hasMemory 16GO

hasFrequency 2.7GHz

hasNetwork 50Gbps

hasLocation Asia Paciﬁc

hasUseCase General Use

hasNonFunctionalVMProperty Conﬁdentiality, Migration, ...

VM21 hasType Small

hasPricePerHour 0.055$

hasMemory 2GO

hasFrequency 2.5GHz

hasNetwork 20Gbps

hasLocation US West

hasUseCase General Use

hasNonFunctionalVMProperty Auto Scaling, Cloud Federation, ...

... ... ...

Table 2: An Extract of PaaS Individuals.

Domain Object Property Range

GoogleAppEngine hasPaaSService Platform11, ...

CloudFoundry hasPaaSService Platform21, ...

... ... ...

Platform11 hasLanguage Java

hasDBMS Oracle

hasPricePerHour 0.7$

hasLocation Asia Paciﬁc

hasNonFunctionalPaaSProperty Security, Integrity, Fault Tolerance

Platform21 hasLanguage Python

hasDBMS MySQL

hasPricePerHour 0.4$

hasLocation Europe

hasNonFunctionalPaaSProperty Cloud Federation, Security, Integrity, Consistency

... ... ...

ated in terms of its qualitative and quantitative as-

pects. The qualitative evaluation treats the logical or

conceptual coherence while the quantitative evalua-

tion is based on some structural criteria (Pradel et al.,

2012).

1. Qualitative Evaluation: Such evaluation tends to

produce a set of properties that reﬂect the respect

for quality criteria. In our work, we are referring

to the following criteria deﬁned by Staab et al. in

(Staab and Studer, 2004) to improve the qualita-

tive evaluation of our proposed ontology:

• Duplication: checks if the elements that can be

deducted needn’t to be explicitly mentioned,

• Disjunction: checks if the class is the conjunc-

tion of disjoint classes,

• Consistency: checks if the deﬁnition of a class

does not lead to a contradiction.

Pellet (Evren et al., 2007), Fact++ (Tsarkov and

Horrocks, 2006) and HermiT (Shearer et al.,

2008) are reasoners which allow the evaluation

of class hierarchy, object property hierarchy,

data property hierarchy, class assertions, ob-

ject property assertions and same individuals.

In this work, to validate our ontology, we

CloudDescriptionOntologyforServiceDiscoveryandSelection

Table 3: An Extract of SaaS Individuals.

Domain Object Property Range

SalesForce hasSaaSService Software11, ...

DoliCloud hasPaaSService Software21, ...

... ... ...

Software11 hasCRM SugarCRM

hasPricePerHour 2.5$

hasLocation Asia Paciﬁc

hasNonFunctionalSaaSProperty Multi Tenancy, Reusability

Software21 hasERP ERP5

hasPricePerHour 4.5$

hasLocation Europe

hasNonFunctionalSaaSProperty Multi Tenancy, Privacy, Availability

... ... ...

Table 4: Users’ Query Examples with Using Ontology and Google.

Query Using the Proposed Ontology Using the Google Engine

Q1 IaaS Provider and hasService value Virtual Machine and IaaS provider

hasPricePerHour max 1$ and hasMemory min 7GO and +price per hour<= 1$

hasNonFunctionalVMProperty value ”Auto Scaling” ˆˆstring +auto scaling

+memory>=7GO

Q2 PaaS Provider and hasService value Platform PaaS provider

and hasPricePerHour max 1.5$ and hasDBMS value Oracle and +Oracle+Java

hasLanguage value Java and +price per hour<= 1.5$

hasNonFunctionalVMProperty value ”Conﬁdentiality” ˆˆstring +conﬁdentiality

Q3 SaaS Provider and hasService value Software and SaaS provider

hasPricePerHour max 4.5$ and hasERP value ERP and +ERP5+multi-tenancy

hasNonFunctionalVMProperty value ”Multi Tenancy” ˆˆstring +price per hour<= 4.5$

choose the Fact++ reasoner which is an open

source, free and available with prot`eg`e. So,

while applying the reasoner, we remark there

exist inconsistency errors. So, we mustn’t

deﬁne the Non

Functional IaaS Property,

the Non Functional PaaS Property and the

Non

Functional SaaS Property classes as disjoint

classes in order to obtain a coherent and valid

Cloud description ontology. Since there exist

QoS individuals which are ISO properties. These

properties describe the non-functional IaaS

services, the non-functional PaaS services and the

non-functional SaaS services.

2. Quantitative Evaluation: The majority of existing

works, evaluate their proposed ontologies while

comparing them with a reference ontology or with

a domain corpus which includes a set of ontolo-

gies and providing a set of similarity measures.

However, in our case, there do not exists a stan-

dard or open ontology to refer to it to it compar-

ing with ours. So, we present users with our on-

tology and ask them to give their Cloud service

queries in order to evaluate its pertinence. In-

deed, we have interviewed three types of users.

The ﬁrst one is an IaaS user who asks for IaaS

providers which offer a virtual machine that has

a price per hour at maximum 1$, a memory ca-

pacity at minimum equal to 7GO and guarantee

the auto-scaling non-functional property (query

Q1). The second is a PaaS user who requests PaaS

providers which offer a platform that has a price

per hour at maximum 1.5$, an Oracle database,

the Java as a language of development and guar-

antee the conﬁdentiality non-functional property

(query Q2). The third user is a SaaS user who

asks for SaaS providers which specify an ERP5

as software, a location at Europe zone and offer

at least the multi-tenancy non functional require-

ment (query Q3). Then, we compare the results

obtained while interrogating the ontology and the

web search engine Google

with users’ queries.

Table 4 describes the queries written in DL-Query

language (Horridge and Drummond, 2008) which

is a plugin for prot`eg`e to make it possible for ex-

http://www.google.com/

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications

Table 5: Query Results with Using Ontology and Google.

Query Ontology Google Correct Correct

Results Results Google Results Ontology Results

Q1 10 13000 1 10

Q2 7 705000 0 7

Q3 5 1280 0 5

Table 6: Results Correctness with Using Ontology and

Google.

Query Google Ontology

Correctness Correctness

Q1 0.008% 100%

Q2 0% 100%

Q3 0% 100%

posure users’ queries using the proposed ontol-

ogy. We take also these examples to ﬁnd results

obtained by the Google web search engine (i.e.,

without using ontology). When interrogating the

ontology and the Google web search engine with

the users’ queries in order to select appropriate

Cloud service provider, we remark that the results

carried out using the Google are not precise and

not adequate. Indeed, we ﬁnd redundant results

and parasite links (i.e., links for blogs, publicity

pages, documents, providers’ catalogs, etc.). In

opposition, while integrating our proposed ontol-

ogy, we obtain quickly and precise results which

satisﬁes all users (i.e., all results are correct or

precise signiﬁes that all founded Cloud providers

offer all functional and non-functional user’s re-

quirements).

Table 5 presents the correctness of both our ontol-

ogy and the Google web engine while discovering

services for the Cloud users. As shown, the ontol-

ogy has a high value of correctness which make it

a relevant and pertinent way for users to select the

appropriate services.

5 CONCLUSION & FUTURE

WORKS

In this work, we proposed a comprehensive ontology

for Cloud service description that covers the three lay-

ers of Cloud models, namely IaaS, PaaS and SaaS.

Our proposed ontology treats the functional and the

non-functional properties of services. In addition, it

improves the interoperability problem among multi-

ples Cloud infrastructures, platforms and applications

within a Cloud federation environment. Furthermore,

it helps users to discover and select appropriate Cloud

services and that is proved by interrogating it with

some examples of users’ queries. The ontology pre-

sented in this paper is mainly based on standards, lit-

erature study, Cloud providers catalogs, analysis and

comparisons of existing Cloud Computing services

ontologies and taxonomies.

As a future work, we aim to improve the eval-

uation of the proposed ontology. First, we plan to

work on populating the ontology with more individ-

uals from real Cloud systems to increase the number

of instances (since the evaluation depends on the in-

troduced population). Second, we investigate to in-

terrogate more number of users’ requests and conse-

quently to determine the precision, the recall and the

inference ability of the ontology. Third, we intend to

compare our proposed ontology with another of the

same ﬁeld. As another perspective, we envisage to ex-

tend our ontology to capture the dependency between

services across the three Cloud models (namely, IaaS,

PaaS and SaaS).

REFERENCES

Alhamad, M., Dillon, T., and Chang, E. (2010). Concep-

tual sla framework for cloud computing. In 4th IEEE

International Conference on Digital Ecosystems and

Technologies, DEST’10, pages 606–610. IEEE.

Behrendt, M., Glaser, B., Kopp, P., Diekmann, R., Bre-

iter, G., Pappe, S., Kreger, H., and Arsanjani, A.

(2011). Introduction and architecture overview ibm

cloud computing reference architecture 2.0. IBM.

Beom, M. Y., Ho, J. S., and Sik, L. J. (2011). Ontology-

based resource management for cloud computing. In

Proceedings of the Third International Conference on

Intelligent Information and Database Systems, ACI-

IDS’11, pages 343–352, Berlin, Heidelberg. Springer-

Verlag.

Bernstein, D. and Vij, D. (2010). Intercloud directory and

exchange protocol detail using xmpp and rdf. In Pro-

ceedings of the 6th World Congress on Services, pages

431–438.

Bhogal, J., Macfarlane, A., and Smith, P. (2007). A re-

view of ontology based query expansion. Inf. Process.

Manage, 43(4):866–886.

Buyya, R., Ranjan, R., and Calheiros, N. (2010). Inter-

cloud: Utility-oriented federation of cloud computing

environments for scaling of application services. In

Proceedings of the 10th International Conference on

CloudDescriptionOntologyforServiceDiscoveryandSelection

Algorithms and Architectures for Parallel Processing

- Volume Part I, ICA3PP’10, pages 13–31. Springer-

Verlag.

Cesar, A. J., Asuncion, G., Lozano, T. A., and Andrade, P.

H. S. (1998). Onto2 agent an ontology based www

broker to select ontologies. In Proceedings of the

Workshop on Applications of Ontologies and Problem

solving Methods at the 13th European Conference on

Artiﬁcial Intelligence.

Cloud, A. E. C. (2013). Amazon ec2. http://aws.

amazon.com/fr/ec2/.

Evren, S., Bijan, P., Cuenca, G. B., Aditya, K., and Yarden,

K. (2007). Pellet: A practical owl-dl reasoner. Web

Semantique, pages 51–53.

Forum, O. G. (2010). Open cloud computing interface

(occi). http://forge.ogf.org/sf/projects/occi-wg.

Ghoneim, A. and Tolba, A. (2014). Cobe framework: Cloud

ontology blackboard environment for enhancing dis-

covery behavior. International Journal on Cloud

Computing: Services and Architecture, 4(4):29–36.

Grozev, N. and Buyya, R. (2012). Inter-cloud architec-

tures and application brokering: Taxonomy and sur-

vey. Software-Practice and Experience, 44:369–390.

Gruber, T. R. (1995). Toward principles for the design of

ontologies used for knowledge sharing. International

Journal of HumanComputer Studies Special issue: the

role of formal ontology in the information technology,

43:907–928.

Henry, L. (2009). Introducing Windows Azure. Apress,

Berkely, CA, USA.

Hoang, H. and Tjoa, A. (2006). The state of the art of

ontology-based query systems a comparison of exist-

ing approaches. In In Proc. of ICOCI06.

Horridge, M. and Drummond, N. (2008). Dlquery.

http://protegewiki.stanford.edu/wiki/DLQuery.

ISO/IEC (2001). ISO/IEC 9126. Software engineering –

Product quality. ISO/IEC.

KPMG (2013). The cloud changing the business ecosys-

tem. http://www.kpmg.com/IN/en/Issues And In-

sights/Thought Leadership/The Cloud Changing the

Business Ecosystem.pdf.

Moscato, F., Aversa, R., and Martino, B. D. (2011). An

analysis of mosaic ontolog y for cloud resources an-

notation. In Proceedings of the Federated Conference

on Computer Science and Information Systems, Fed-

CSIS’11, pages 973–980.

NIST (2013). Nist Cloud Computing Standards Roadmap.

CreateSpace Independent Publishing Platform.

Parhi, M., Pattanayak, B., and Patra, M. (2014). A multi-

agent-based framework for cloud service description

and discovery using ontology. In InProceedings of

Intelligent Computing, Communication and Devices,

volume 1 of ICCD 2014, pages 337–348. Springer In-

dia.

Pradel, C., Hernandez, N., Kamel, M., and Rothenburger,

B. (2012). Une ontologie du cin´ema pour ´evaluer

les applications du web s´emantique. In In Atelier

Ontologies et Jeux de Donn´ees pour ´evaluer le web

s´emantique, IC’2012.

Rekik, M., Boukadi, K., and Ben-Abdallah, H. (2014). A

context based scheduling approach for adaptive busi-

ness process in the cloud. In IEEE 7th International

Conference on Cloud Computing, CLOUD’2014,

pages 948–951. IEEE.

Rochwerger, B., Breitgand, D., Levy, E., Galis, A., Nagin,

K., Llorente, M., andY. Wolfsthal, R. M., Elmroth,

E., Caceres, J., Ben-Yehuda, M., Emmerich, W., and

Galan, F. (2009). The reservoir model and architec-

ture for open federated cloud computing. IBM J. Res.

Dev., 53(4):535–545.

Russell, J. and Cohn, R. (2012). Gogrid. Book on Demand.

Ruthven, I. and Lalmas, M. (2003). A survey on the use

of relevance feedback for information access systems.

Knowl. Eng. Rev., 18(2):95–145.

Shearer, R., Motik, B., and Horrocks, I. (2008). Hermit: A

highly-efﬁcient owl reasoner. In Proc. of the 5th Int.

Workshop on OWL: Experiences and Directions.

Staab, S. and Studer, R., editors (2004). Handbook on

Ontologies. International Handbooks on Information

Systems. Springer.

Tsarkov, D. and Horrocks, I. (2006). Fact++ description

logic reasoner: System description. In Proceedings

of the International Joint Confererence on Automated

Reasoning, pages 292–297.

Youseff, L., Butrico, M., and Silva, D. D. (2008). Towards

a uniﬁed ontology of cloud computing. In in Proceed-

ings of the Grid Computing Environments Workshop,

GCE08, pages 1–10.

Zhang, M., Ranjan, R., Haller, A., Georgakopoulos, D.,

Menzel, M., and Nepal, S. (2012). An ontology

based system for cloud infrastructure services discov-

ery. CoRR.

ICSOFT-EA2015-10thInternationalConferenceonSoftwareEngineeringandApplications