A TAXONOMY MODEL FOR CLOUD COMPUTING SERVICES

Nelson M. Gonzalez, Charles C. Miers, Fernando F. Red

ıgolo, Marcos Simpl

ıcio,

Tereza C. M. B. Carvalho

Laboratory of Computer Architecture and Networks, University of S

ao Paulo

Av. Prof. Luciano Gualberto, Trav. 3 - Nr. 158 - C1 50a, S

ao Paulo, Brazil

Mats N

aslund

, Makan Pourzandi

Ericsson Research, Stockholm, Sweden

Ericsson Research, Ville Mont-Royal, Canada

Keywords:

Cloud Computing, SPI, Taxonomy.

Abstract:

The continuous development of cloud computing is in evidence in several academic and non-academic re-

searches. However, the relative youth of this ﬁeld has produced several distinct deﬁnitions and taxonomies

regarding the concept of cloud computing, as well as the classiﬁcation and organization of such services. The

appearance of commercial cloud solutions in this context with no ﬁrmly established standards only compli-

cates the matter, making it difﬁcult to determine how solutions should be technically identiﬁed and qualiﬁed.

Therefore, with the growing complexity of the area, identifying, clarifying and classifying cloud services are

essential steps to understand their organization, purpose and interaction with other services. With this goal in

mind, this article presents a study on existing concepts and taxonomies, and then harmonizes these approaches

in an extensible taxonomy model for cloud computing services. More speciﬁcally, this proposal builds on the

SPI (Software, Platform, and Infrastructure) taxonomy created by NIST (National Institute of Standards and

Technology), creating a hierarchical organization that groups different services according to their character-

istics; the result is a taxonomy model that allows ﬁner-grained analyses to be performed, while essentially

keeping the simplicity of SPI itself. Finally, we present a speciﬁc instance of this model focused on existing

and representative cloud services.

1 INTRODUCTION

The concept of cloud computing is the center of atten-

tion in several academic and non-academic research

efforts. Cloud computing is sometimes considered

the natural evolution of Internet (Velte et al., 2009),

offering services able to replace hardware and soft-

ware purchased and maintained by companies and

customers at their own risk and expense. The ﬁrst

main advantage of this approach is the clearer dis-

tinction between the business itself (e.g., selling mul-

timedia content) and the tools required by the busi-

ness (Coombe, 2009) (e.g., datacenters and computer

programs). Hence, cloud services allow customers to

focus on their projects, leaving all unrelated aspects

to the service provider.

Since cloud computing is a recent and evolving

concept, it still lacks a precise classiﬁcation. Indeed,

even core aspects such as the possible structures, ar-

chitectures and application scenarios still remain un-

der continuous discussion, which is reasonable to ex-

pect for a technology whose impact and grandiosity is

compared to the development of Internet itself (Arm-

brust et al., 2009). In this context, it is important to

understand how cloud services interact with other ser-

vices, as well as the mechanisms that can (or must) be

used in order to assure a proper communication be-

tween them. This task, on the other hand, requires

a robust model capable of classifying the services

according to their types, which has lead to the cre-

ation of several taxonomies with this purpose (Jaekel

and Luhn, 2009). One of the most accepted and

adopted models for building such taxonomies is the

SPI model (Mell and Grance, 2009), which divides

services into three main categories: Software as a Ser-

vice (SaaS), Platform as a Service (PaaS), and Infras-

tructure as a Service (IaaS).

In this paper, we present an enhanced taxonomy

M. Gonzalez N., C. Miers C., F. Redígolo F., Simplício M., C. M. B. Carvalho T., Näslund M. and Pourzandi M..

A TAXONOMY MODEL FOR CLOUD COMPUTING SERVICES.

DOI: 10.5220/0003384800560065

In Proceedings of the 1st International Conference on Cloud Computing and Services Science (CLOSER-2011), pages 56-65

ISBN: 978-989-8425-52-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

model for cloud computing solutions, which employs

an hierarchical structure to harmonize the concise

approach adopted by the SPI model with the ﬁner-

grained structure from other taxonomies, namely

those presented in (Youseff et al., 2008; Johnston,

2010; Rimal et al., 2009a; Laird, 2009; Linthicum,

2009). Since these taxonomies were designed specif-

ically to identify how services can be compared and

combined (Youseff et al., 2008), the proposed model

allows a more involving organization of cloud solu-

tions. This model can thus be seen as an enhanced

version of (rather than a substitution for) the SPI

model, offering relevant criteria for classifying exist-

ing services and also allowing reusability and com-

posability of new ones. Moreover, in order to illus-

trate these capabilities, we also present a speciﬁc in-

stance of the proposed taxonomy model, focusing on

existing cloud services. Both the model and this spe-

ciﬁc instance aim to guide further studies on cloud

computing and the lifecycle of related services, focus-

ing on how such services may be created and conﬁg-

ured to perform different and complementary tasks.

The rest of this document is organized as fol-

lows: Section 2 provides an overview of cloud com-

puting based on ofﬁcial deﬁnitions conceived by

NIST (National Institute of Standards and Technol-

ogy) (Mell and Grance, 2009) and ENISA (European

Network and Information Security Agency), empha-

sizing the key characteristics of cloud services. Sec-

tion 3 presents and brieﬂy describes some relevant

taxonomies that compose the basis of our proposal.

These taxonomies are then compared in section 4,

which explores points to be reproduced and main

characteristics taken into account in the proposed tax-

onomy model, presented in section 5. The related

work is covered in section 6. Section 7 provides our

ﬁnal considerations, comparing the proposed model

with existing approaches and also presenting sugges-

tion of future work in the area.

2 CLOUD COMPUTING

OVERVIEW

Cloud computing solutions are being developed and

becoming available for ﬁnal users despite the lack

of a clear classiﬁcation, and while new features are

still under research (IDC, 2010). A recurring issue

in this scenario is the free and unrestrained creation

of deﬁnitions and classiﬁcation categories by service

providers, which are commonly attempting to pro-

mote their solutions’ features rather than following a

technical perspective (Willis, 2009). Probably moti-

vated by this concern, both NIST and the European

Community started the process of standardizing the

key aspects of cloud computing, aiming to unify the

existent concepts and to simplify the task of identify-

ing the main purposes of each cloud solution (Hoover,

2009).

NIST efforts to standardize the basic cloud com-

puting concepts are associated to the increasing use

of cloud computing solutions by USA governmental

agencies. Therefore, most of the institute’s work fo-

cus on USA government requirements for cloud com-

puting, both from the technical and legal points of

view. According to NIST, cloud computing can be

deﬁned as follows (Mell and Grance, 2009):

“Cloud computing is a model for enabling

convenient, on-demand network access to a

shared pool of conﬁgurable computing re-

sources (e.g., networks, servers, storage, ap-

plications, and services) that can be rapidly

provisioned and released with minimal man-

agement effort or service provider interac-

tion. This cloud model promotes availability

and is composed of ﬁve essential character-

istics, three service models, and four deploy-

ment models.”

The main characteristics of cloud computing as

deﬁned by NIST are (Mell and Grance, 2009):

• On-demand self-service: The resources provided

are easily scaled, so the customer is able to allo-

cate or free them automatically whenever needed,

without human interaction with providers;

• Broad network access: Services are available

through Internet using standardized interfaces

(such as browsers or remote console control), be-

ing accessible from any device able to connect and

use these interfaces;

• Resource pooling: Provider infrastructure is or-

ganized to offer services for multiple customers

using the concept of multi-tenancy, according to

which resources are managed according to the de-

mand. The resource management is transparent to

users.

• Rapid elasticity: Cloud capabilities are promptly

and easily scaled to satisfy the customers’ needs,

in such a manner that the cloud resources appear

to be unlimited from the users’ point of view; and

• Measured service: Services are automatically

controlled and optimized to offer metering capa-

bilities according to the type of service.

In comparison, ENISA (ENISA, 2009) deﬁnes

cloud computing as an on-demand service model for

IT provision, based on virtualization and distributed

A TAXONOMY MODEL FOR CLOUD COMPUTING SERVICES

computing technologies. According to this agency,

the key characteristics of cloud computing are:

• Highly abstracted resources: Virtualized infras-

tructures are the core of cloud computing, either

by directly offering a virtualized platform or de-

livering services based on these structures;

• Scalability and ﬂexibility: Clouds are highly scal-

able, provisioning customers with the required re-

sources; they are also ﬂexible enough to conform

itself to the customers’ needs;

• Instantaneous provisioning: The resources appear

to the customers as unlimited and promptly provi-

sioned as necessary;

• Shared resources: The provider infrastructure

(including hardware, database and memory) is

shared by customers. This also has security im-

plications, as the logical layers can be separated

but the hardware cannot;

• Service on demand with “pay as you go” billing

system: The elasticity of the cloud enables the

precise deﬁnition of what resources the customer

needs, and when they are needed; and

• Programmatic management: Web services pro-

gramming interfaces allow the management of the

cloud infrastructure and also the integration be-

tween local software and cloud resources.

We note that both deﬁnitions (and also oth-

ers (CPNI, 2010; CSA, 2009)) highlight the “pay as

you go” feature for highly abstracted resources, pro-

vided by a same shared infrastructure that, as such,

must achieve a minimum level of scalability and ﬂex-

ibility in order to adapt itself to the demand from mul-

tiple tenants.

3 EXISTING TAXONOMIES

Cloud computing comprises several types of services,

from virtualized storage to development platforms.

This complexity lead to the development of many tax-

onomies aiming to classify the different services and

group those with similar characteristics, which is cru-

cial for a more systematic study of such multifaceted

area. Some of these taxonomies are very concise,

while others create different categories with suppos-

edly heterogeneous characteristics, as discussed in the

following.

3.1 NIST’s SPI Model

As brieﬂy discussed in section 1, NIST’s SPI

model (Mell and Grance, 2009) is currently one of

the most well-accepted taxonomies for cloud com-

puting (CSA, 2009; Marks and Lozano, 2010). Its

main objective is to encompass all various cloud ap-

proaches (models, vendors and market niches) in or-

der to standardize concepts. In this manner, the SPI

model classiﬁes the cloud solutions in three cate-

gories:

• Software as a Service (SaaS): This category is

related to cloud services that are developed, de-

ployed, run and maintained by the provider and

which are accessible through a Web browser. The

customer does not manage or control the under-

lying infrastructure which supplies the required

resources while securing the assets involved and

guaranteeing service level agreements;

• Platform as a Service (PaaS): Services that pro-

vide a platform to develop and deploy web appli-

cations. The customer again does not have control

over the cloud infrastructure but is able to directly

access and use its resources, commonly made

available through the use of specialized APIs (Ap-

plication Programming Interfaces); and

• Infrastructure as a Service (IaaS): Cloud solutions

able to offer infrastructural utilities and tools,

such as processing, storage, virtual machines or

other computing basic resources. The customer

has considerable control over the infrastructure,

even though there is no direct access to the physi-

cal hardware itself;

This taxonomy presents a succinct but compre-

hensive classiﬁcation of cloud services abstracting

three main possibilities of exploring them: to offer

online browser-ready applications running over a vir-

tualized infrastructure (SaaS); to provide platforms

for the deployment of applications which use the un-

derlying resources from the provider’s infrastructure

(PaaS); or to deliver the virtualized infrastructure it-

self (IaaS).

3.2 Youseff’s 5-Layer Ontology

Youseff et al. (Youseff et al., 2008) conducted a re-

search on cloud computing aiming at a ﬁnal and con-

solidated taxonomy to be used for general purposes.

The proposed taxonomy features ﬁve layers contain-

ing altogether six different categories depicted in Fig-

ure 1 and described as follows:

• Software as a Service (SaaS): Aggregates all

software-oriented solutions, e.g., Salesforce CRM

(Salesforce, 2010b), Google Apps (Google,

2010d);

• Platform as a service (PaaS): Application de-

velopment platforms, e.g. Google App Engine

CLOSER 2011 - International Conference on Cloud Computing and Services Science

Figure 1: Youseff’s 5-layer ontology (Youseff et al., 2008).

(Google, 2010a), Salesforce Apex (Salesforce,

2010a), Hadoop (Borthakur, 2007), Yahoo Pig!

(Olston et al., 2008);

• Infrastructure as a Service (IaaS): Computational

resources that can be moved to the cloud, mainly

represented by virtualization, e.g. Amazon EC2

(Amazon, 2010b), Enomalism (Enomaly, 2010),

Eucalyptus (Nurmi et al., 2009), OpenNebula

(Font

an et al., 2008);

• Data as a Service (DaaS): Refers to remotely

managed solutions for data storage, e.g., Google

FileSystem (Google, 2010b), Bayou (Demers

et al., 1994), Dynamo (DeCandia et al., 2007),

Amazon S3 (Amazon, 2010e), EMC (EMC,

2010);

• Communication as a Service (CaaS): Solutions

providing service-oriented communication for

cloud users, including VoIP systems, instant

messaging, and audio/video conferencing, e.g.

Microsoft Connected Services Framework (Mi-

crosoft, 2010a; Hofstader, 2007); and

• Hardware as a Service (HaaS): Refers to services

where the provider operates, manages and up-

grades the hardware used by the customer, e.g.,

Morgan Stanley IBM utility computing (Hines,

2004) PXE (Johnston, 1999), UBL (UBOOT,

2010), IBM Kittyhawk (Appavoo et al., 2008).

The ﬁve layers abstract computing environments,

starting from the application (top layer), going

through the environment (which provides the re-

sources to the application to run), software infrastruc-

ture (the resources that are enabled by the operating

system), kernel and ﬁnally hardware. These layers are

used to identify possible categories for cloud services,

resulting in the taxonomy presented.

3.3 Sam Johnston’s 6-Layer Stack

Judging that NIST’s classiﬁcation based on three lay-

ers/categories was oversimpliﬁed, Johnston (John-

ston, 2010) created a cloud computing taxonomy fo-

cusing on the software for the ﬁnal users. The result-

ing taxonomy organizes the cloud ecosystem in six

layers, illustrated in Figure 2 and detailed as follows:

• Clients: Computer hardware and/or software that

rely on the cloud to deliver services and applica-

tions, e.g., browsers in general, cloud operating

systems, Cirtas (Cirtas, 2010), g-Eclipse (Gjer-

mundrod et al., 2008), StorSimple (StorSimple,

2010);

• Services: Web services that allow the direct in-

teraction between customers and providers’ solu-

tions;

• Application: Solutions that eliminate the need of

installing or running software on local machines,

e.g., Google Apps (Google, 2010d), Zoho (Zoho,

2010);

• Platform: Computing platform to deploy applica-

tions without having to purchase the required in-

frastructure, e.g., Force.com (SalesForce, 2010),

Google app Engine (Google, 2010a), Heroku

(Heroku, 2010), OrangeScape (OrangeScape,

2010), Rackspace (Rackspace, 2010);

• Storage: Delivering storage as a service, either

in the form of raw storage or database utili-

ties, e.g., Amazon S3 (Amazon, 2010e), Amazon

SimpleDB (Amazon, 2010f), Dropbox (Dropbox,

2010), Evernote (Evernote, 2010), Mozy (De-

cho, 2010), Windows Live Skydrive (Microsoft,

2010b); and

• Infrastructure: Services that offer virtualization

solutions for customers, allowing direct access

to the cloud infrastructure, e.g., Amazon EBS

(Amazon, 2010a), Amazon MI (Amazon, 2010c),

AppNexus (AppNexus, 2010), Cloudera (Cloud-

era, 2010), ElasticHosts (ElasticHosts, 2010),

GoGrid (GoGrid, 2010), Hadoop (Borthakur,

2007), RightScale (RightScale, 2010).

Figure 2: Sam Johnston’s 6-layer stack.

The resulting stack shares many similarities with

the one used by Youseff et al., but includes the

A TAXONOMY MODEL FOR CLOUD COMPUTING SERVICES

“Clients” layer, which represents elements that effec-

tively use the services from the adjacent layer.

The same taxonomy can also be presented as in

Figure 3, which shows the categories and the exam-

ples used by the authors to contextualize them.

Figure 3: Sam Johnston’s 6-layer stack.

3.4 Dave Linthicum’s 10-Layer

Taxonomy

Linthicum’s taxonomy (Linthicum, 2009) aggregates

all possible services into ten categories on the same

level (see Figure 4).

Figure 4: Dave Linthicum’s 10-layer taxonomy.

These categories are described as follows:

• Storage as a Service: Ability to leverage stor-

age handled by the customer, so it becomes phys-

ically remote but logically local, e.g., Amazon

S3 (Amazon, 2010e), Box.net (Box.net, 2010),

Google Base (Google, 2010c);

• Database as a Service: Follows the same principle

of Storage, but focuses exclusively on databases,

offering the space need for storing the data and

also interfaces to use the database tools, e.g.,

Amazon SimpleDB (Amazon, 2010f), Trackvia

(Trackvia, 2010);

• Information as a Service: Offers APIs and inter-

faces to access remotely hosted information, such

as stock prices, address validation, credit report-

ing or meteorology;

• Process as a Service: Provides structures able to

bind other resources altogether, creating business

processes easily changeable, e.g., Appian Any-

where (Appian, 2010), Itensil (Itensil, 2010);

• Application as a Service: Software accessible

through browser, e.g., Salesforce CRM (Sales-

force, 2010b), Google Apps (Google, 2010d);

• Platform as a Service: Includes the application

development platform, interfaces to access the

cloud infrastructure (including database utilities),

raw storage and testing, e.g., Google App Engine

(Google, 2010a), Heroku (Heroku, 2010);

• Integration as a Service: An integration stack that

includes application interfaces, ﬂow control and

integration designs, allowing the intercommuni-

cation between different programs or processes,

e.g., Amazon SQS (Amazon, 2010d);

• Security as a Service: Mainly related to identity

management, e.g., Ping Identity (Identity, 2010);

• Management as a Service: On-demand services

that provide resources and structure to manage

other cloud services in terms of topology, resource

utilization, virtualization and task management,

e.g., RightScale (RightScale, 2010); and

• Testing as a Service: Services able to test

other cloud services, eliminating performance is-

sues and bugs during development, deployment

and production stage, e.g., SOASTA (SOASTA,

2010).

This taxonomy introduces many categories to di-

vide the cloud ecosystem. However, it lacks some

important aspects, as the IaaS set of services is not

entirely covered (there is not a clear category for vir-

tualization services) and there are no clear interrela-

tionships between different categories.

CLOSER 2011 - International Conference on Cloud Computing and Services Science

4 COMPARISON AND

LIMITATIONS OF EXISTING

PROPOSALS

The ﬁrst point worth noticing when comparing the

four taxonomies previously presented refers to the

number of categories adopted: the most succinct is

the SPI model with only three categories, followed

by Youseff’s ﬁve-layers ontology, then by Johnston’s

six-layers stack and ﬁnally by Linthicum’s taxonomy

with ten different categories. Despite this difference,

they all follow a similar organization and assume a

linear or ﬂat structure. Hence, it is plausible to use

the SPI model as reference to compare the other tax-

onomies, as depicted in Table 1. Even though the

SPI model was not necessarily used by the other tax-

onomies presented, this table helps to identify the

main points where each layer can be subdivided.

Table 1: Taxonomies comparison using SPI model as refer-

ence.

Taxonomy SaaS PaaS IaaS

Youseff’s

Software Platform

Comp. resources

5-layer

Storage

Communications

Johnston’s Services

Platform

Storage

6-layer Application Infrastructure

Linthicum’s

Application

10-layer

Testing Platform

Storage

Security Integration

Database

Information Management

Process

One of the most important requirements when

specifying cloud taxonomies is to permit an easier

identiﬁcation of interrelationships between different

services; this allows the services to be studied sepa-

rately, as well as the creation of new services by com-

posing simpler ones.

The straightforward approach proposed in the SPI

model is helpful when a simple and direct classiﬁ-

cation is needed, but encumbers the identiﬁcation of

interrelationships between services. In comparison,

the other models described have more categories in a

ﬂat distribution; however, they rely basically on exist-

ing services to create the categories, in such a man-

ner that the categories themselves usually do not have

clear connections. One of the main disadvantages of

this more empiric approach is that the resulting cat-

egories are likely unable to fully or properly cover

future services, or even existing services that were

not considered when the taxonomy was ﬁrst designed.

One example is the Amazon EC2 (Amazon, 2010b),

which does not ﬁt in any of Linthicum’s taxonomy

(Linthicum, 2009) categories: even though it is a stor-

age service, it also offers virtualization and database

features.

We can conclude that, despite their importance,

the above-mentioned taxonomies are either too broad

– not providing the granularity needed for deeper and

more complex comparisons between distinct services

– or too narrow – lacking the ﬂexibility for adding

new, more complex services, – and do not offer an ex-

plicit connection between the different categories. In

this context, the development of a more ﬂexible and

comprehensive approach gains interest.

5 PROPOSED TAXONOMY

The proposed taxonomy model builds on the widely

accepted SPI, but adopts an hierarchical organization

rather than a ﬂat one. This approach leads to a more

structured taxonomy, not only conserving the simplic-

ity of the SPI model in its higher level, but also en-

abling a deeper analysis of cloud services in its lower

levels, as aimed by the other taxonomies described in

section 3. Indeed, one of the most interesting features

of the SPI model is its ability to offer an overview

of services and to initially categorize them; the pro-

posed taxonomy model extends this property with a

ﬁner-grained structure, enabling the development of

the more sophisticated analyses that are not possible

with the SPI model alone.

The result is a tree-like structure that aggregates

cloud solutions, favoring the identiﬁcation of simi-

larities, dependencies and complementary points be-

tween these services. Figure 5 illustrates the proposed

model, showing a taxonomy based on existing ser-

vices. The speciﬁc categories depicted are:

• Software:

– e-Commerce: Services that enable the creation

of online stores and catalogs for local physi-

cal products, and application stores for mobile

platforms;

– Software management: Services that commu-

nicate with other software (hosted or not in a

cloud) with management purposes; such as a

hosted web application or an antivirus server.

This category can be further divided into mon-

itoring services (e.g., for checking service up-

time, performance, security or SLA compli-

ance), controlling (for creating, starting and

stopping other services) and integrating (for

data synchronization and conversion);

– Financial management: Financial transactions,

such as payment and billing services;

A TAXONOMY MODEL FOR CLOUD COMPUTING SERVICES

Figure 5: A service-based instance of the proposed taxon-

omy model.

– Communication: General-purpose types of

communication tied to an online event (e.g.,

VoIP, webcasts, webinars and web conferenc-

ing solutions);

– Ofﬁce tools: Word processors, spreadsheets,

presentation software, ﬁle conversion, calen-

dars, etc.;

– Entertainment: Online games, video/audio

streaming, which demand speciﬁc QoS require-

ments; and

– Business management: Corporate applications

such as CRM (Customer Relationship Manage-

ment) and BPM (Business Process Manage-

ment).

• Platform:

– Developing: Hosted environment for soft-

ware development, like IDEs (Integrated De-

velopment Environment), versioning and bug-

tracking systems;

– Deploying: Services that offer an environment

to deploy applications, from static to dynamic

web pages and web services; and

– Testing: Solutions for automated tests of new

applications.

• Infrastructure:

– Processing: Virtualized processing resources,

e.g., virtual machines execution;

– Storage: Virtualized storage, including

database, ﬁle storage, and disk storage; and

– Networking: Virtualized networks and services

on top of an existing network infrastructure,

such as wireless clouds.

The proposed format enables not only to clas-

sify a service using speciﬁc categories (represented

by the leaves), but also using more generic levels,

which are especially recommended when a service

is very complex or when it aggregates different so-

lutions. Some Amazon web services are interesting

examples to contextualize both situations. The Ama-

zon S3 (Amazon, 2010e) is a storage service, which

is initially classiﬁed as an IaaS; using the proposed

model, this service is classiﬁed as a Infrastructure-

Storage. On the other hand, Amazon EC2 (Ama-

zon, 2010b), which is related to virtualization, offers

a complete platform to run software over a virtualized

infrastructure; it is classiﬁed as an Infrastructure ser-

vice, because it offers resources related to its leaves.

In this case it is interesting to note that both services

would be simply classiﬁed as IaaS by the SPI model

– which is correct and adequate if a simple analysis

is required, but insufﬁcient if one requires a compar-

ison between various IaaS providers, each of which

offering different solutions.

We emphasize, however, that the categories enu-

merated above are not necessarily exhaustive. In-

stead, the taxonomy instance depicted in Figure 5

is better regarded as an application of the proposed

model for common cloud services; as such, this tax-

onomy can be easily extended without losing its main

characteristics, as long as the SPI root is kept un-

touched. This extensibility feature also enables a clear

evolution path, similar to the way networking stan-

dards are created and maintained by standardization

entities such as IEEE and IETF (for instance, using

tags to identify the taxonomy version).

It is important to emphasize that the combinations

between services may be very complex. For example,

while using Heroku to develop a dynamic website for

an enterprise, the service consumer may decide to in-

clude Netsuite Ecommerce to integrate local products

to an online store, to use Amazon S3 or SimpleDB

to store any information related to business processes,

customers and commercial transactions, and ﬁnally to

adopt SOASTA to improve the website performance.

This example illustrates that classifying services into

more complex categories whilst keeping a solid foun-

dation enables a swift and useful identiﬁcation of po-

tential services that, when combined, can create a

complete toolkit for developing the desired business.

Table 2 shows the classiﬁcation of some existing

services according to this taxonomy.

CLOSER 2011 - International Conference on Cloud Computing and Services Science

Table 2: Classiﬁcation of existing services using proposed

taxonomy.

Service Main category Subdivisions

Amazon EC2 (Amazon, 2010b) Infrastructure -

Amazon S3 (Amazon, 2010e) Infrastructure Storage

Amazon SimpleDB (Amazon, 2010f) Infrastructure Storage – Database

Amazon SQS (Amazon, 2010d) Sofware Software management

– Integration

Appian Anywhere (Appian, 2010) Software Business management

– BPM

Boomi (Boomi, 2010) Software Software management

– Integration

Box.net (Box.net, 2010) Infrastructure Storage

Elastra (Das et al., 2010) Software Software management

EMC SMS (EMC, 2010) Infrastructure Storage

Enomaly ECP (Enomaly, 2010) Infrastructure -

Google App Engine (Google, 2010d) Platform Deploying

Heroku (Heroku, 2010) Platform Developing, Deploying

Netsuite Apps (Netsuite, 2010) Software Financial management

OpSource Connect (OpSource, 2010) Software Software management

– Integration

Oracle OnDemand (Oracle, 2010) Software Business management

– CRM

RightScale (RightScale, 2010) Software Cloud management

rPath (rPath, 2010) Software Cloud management

Salesforce CRM (Salesforce, 2010b) Software CRM

SOASTA (SOASTA, 2010) Platform Testing

Trackvia (Trackvia, 2010) Infrastructure Database

6 RELATED WORK

In addition to the taxonomies for cloud computing

services discussed in Section 3, there are also other

taxonomies that aim to identify further criteria to clas-

sify those services.

In (Prodan and Ostermann, 2009), Prodan and

Ostermann focus on the growth of cloud solutions

and the constant need of fast and scalable resources

for solving complex and demanding problems from

both scientiﬁc and business applications The authors

provide an organization of cloud computing envi-

ronments based on eight main aspects: the service

type, how resources are deployed, the hardware, run-

time tuning, security concerns, the business model

adopted, any middleware involved and performance

requirements. The organization adopted is very sim-

ilar to the SPI model, having as main difference the

addition of a fourth category denominated special-

ized services, which includes web and ﬁle hosting.

The cloud solutions are then categorized according to

several different criteria, such as environments, ser-

vice type, deployment, business model, Application

Programming Interface (API), and performance. As

a result, the proposed classiﬁcation ends up resem-

bling a comparison/benchmark of the listed cloud so-

lutions rather than a taxonomy, requiring a consider-

able amount of information on how the solution was

built and a deep understanding on its speciﬁcation and

requirements. In comparison, the taxonomy model

and its instance proposed in this paper provides a

more straightforward classiﬁcation, letting the bench-

mark comparison to be veriﬁed on demand and ac-

cording to the consumer’s desire (in terms of require-

ments and criteria).

Rimal et al. (Rimal et al., 2009b), on the other

hand, highlight the acceptance of pay-for-use models

and of the everything-as-a-Service approach. They

state that the taxonomies currently available were cre-

ated from the perspective of vendors, not from the ac-

tual consumers of the cloud services. With this issue

in mind, the authors deﬁne many criteria to classify

cloud solutions, such as cloud architecture, virtual-

ization management, services, fault tolerance, secu-

rity, and other issues. The proposed taxonomy is also

based on the SPI model, but includes an additional

layer denominated Hardware as a Service (HaaS), lo-

cated between the Paas and IaaS categories. This new

category is, however, very similar to the IaaS layer

from the original SPI Model, while the then redeﬁned

IaaS includes functionalities such as payment, which

is usually classiﬁed in the SaaS category. The taxon-

omy model, on the other hand, extends the SPI model

in a more elegant manner; moreover, the presented in-

stance focus only on the services category, since it is

related only to the taxonomy of cloud services.

7 DISCUSSION AND

CONSIDERATIONS

Cloud computing is a powerful technology, enabling

productive and constructive use of computational re-

sources, easing software development and distribut-

ing applications through the Web. In this context,

it is important to have a concise yet comprehensive

taxonomy for organizing and classifying cloud solu-

tions, which is many times provided by the solid and

well established SPI model. One shortcoming of this

model, however, is that it lacks of a clear connection

between services inside each category.

In this paper, we address this issue by proposing

an enhanced taxonomy model that extends the SPI

model by means of an hierarchical structure. In a ﬁrst

level, this new model conserves the simplicity of SPI,

A TAXONOMY MODEL FOR CLOUD COMPUTING SERVICES

allowing an easy identiﬁcation of the overall charac-

teristics of cloud services. Then, as it goes deeper in

the tree-like classiﬁcation, it enables more sophisti-

cated analyses, which includes the capability of de-

termining existing correlations between distinct ser-

vices. These correlations allow the identiﬁcation of

complementary points that can be explored in order

to offer a complete set of virtualized tools (from soft-

ware to hardware), so the cloud customer is able to

focus on projects rather than the required infrastruc-

ture. Therefore, the aim of the proposed model is an

extension of (rather than a replacement for) the SPI

model, empowering the latter with an organization

that allows a ﬁner-grained analysis to be performed

whenever needed.

Another essential feature of the proposed model

is the extensibility from the SPI root. Hence, its in-

stances should not be regarded as ﬁxed and absolute

taxonomies, but as evolving structures that can fol-

low the cloud computing evolution itself. This also

means that future studies on criteria and other func-

tional aspects of the cloud may very well improve the

proposed model, organizing the services in order to

facilitate the development of more sophisticated cloud

computing solutions.

ACKNOWLEDGEMENTS

This work was supported by the Research and De-

velopment Centre, Ericsson Telecomunicac¸

oes S.A.,

Brazil.

REFERENCES

Amazon (2010a). Amazon elastic block storage. http://

aws. amazon.com/ebs/.

Amazon (2010b). Amazon elastic compute cloud. http://

aws.amazon.com/ec2/.

Amazon (2010c). Amazon machine images. http://

aws.amazon.com/amis/.

Amazon (2010d). Amazon simple queue service. http://

aws.amazon.com/sqs/.

Amazon (2010e). Amazon simple storage service. http://

aws.amazon.com/s3/.

Amazon (2010f). Amazon simpledb. http://

aws.amazon.com/simpledb/.

Appavoo, J., Uhlig, V., and Waterland, A. (2008). Project

Kittyhawk: building a global-scale computer: Blue

Gene/P as a generic computing platform. SIGOPS

Oper. Syst. Rev., 42:77–84.

Appian (2010). Appian anywhere. http://www.appian.com/

bpm-saas.jsp.

AppNexus (2010). Appnexus cloud computing. http://

www.appnexus.com/.

Armbrust, M., Fox, A., Grifﬁth, R., Joseph, A., Katz, R.,

Konwinski, A., Lee, G., Patterson, D., Rabkin, A.,

Stoica, I., and Zaharia, M. (2009). Above the clouds:

A berkeley view of cloud computing. Technical re-

port.

Boomi (2010). http://www.boomi.com/solutions.

Borthakur, D. (2007). The hadoop distributed ﬁle system:

Architecture and design. http://hadoop.apache.org/.

Box.net (2010). Box. http://www.box.net/solutions.

Cirtas (2010). Cirtas bluejet. http://www.cirtas.com/.

Cloudera (2010). Cloudera. http://www.cloudera.com/.

Coombe, B. (2009). Cloud computing - overview, advan-

tages and challenges for enterprise deployment. Tech-

nical report, Bechtel Corporation.

CPNI (2010). Information security brieﬁng 01/2010 - cloud

computing. Technical report, CPNI - Centre for Pro-

tection of National Infrastructure.

CSA (2009). Security guidance for critical areas of focus

in cloud computing. Technical report, CSA - Cloud

Security Alliance.

Das, S., Agrawal, D., and Abbadi, A. E. (2010). Elas-

TraS: An elastic, scalable, and self managing transac-

tional database for the cloud. Technical report, UCSB.

http://www.cs.ucsb.edu/research/tech

reports/.

DeCandia, G., Hastorun, D., Jampani, M., Kakulapati,

G., Lakshman, A., Pilchin, A., Sivasubramanian, S.,

Vosshall, P., and Vogels, W. (2007). Dynamo: Ama-

zon’s highly available key-value store. In Proc. of

ACM Symposium on Operating Systems Principles,

pages 205–220. ACM.

Decho (2010). Mozy. http://mozy.com/.

Demers, A., Petersen, K., Spreitzer, M., Terry, D., Theimer,

M., and Welch, B. (1994). The Bayou architecture:

Support for data sharing among mobile users. In Proc.

of the Workshop on Mobile Computing Systems and

Applications, pages 2–7.

Dropbox (2010). Dropbox. http://www.dropbox.com/.

ElasticHosts (2010). Elastichosts – ﬂexible servers in the

cloud. http://www.elastichosts.com/.

EMC (2010). Emc storage managed services. http://

www. emc.com/.

ENISA (2009). Beneﬁts, risks and recommendations for

information security. Technical report, ENISA - Eu-

ropean Network and Information Security Agency.

Enomaly (2010). Enomalism elastic computing infrastruc-

ture. http://www.enomalism.com/.

Evernote (2010). Evernote. http://www.evernote.com/.

Font

an, J., V

azquez, T., Gonzalez, L., Montero, R., and

Llorente, I. (2008). Opennebula: The open source vir-

tual machine manager for cluster computing. In Open

Source Grid and Cluster Software Conference - Book

of Abstracts. http://www.opennebula.org/.

CLOSER 2011 - International Conference on Cloud Computing and Services Science

Gjermundrod, H., Dikaiakos, M., Stumpert, M., Wol-

niewicz, P., and Kornmayer, H. (2008). g-Eclipse -

an integrated framework to access and maintain grid

resources. In Proc. of the 2008 9th IEEE/ACM Inter-

national Conference on Grid Computing, GRID’08,

pages 57–64, Washington, DC, USA. IEEE Computer

Society.

GoGrid (2010). http://www.gogrid.com/.

Google (2010a). http://code.google.com/appengine/.

Google (2010b). http://labs.google.com/papers/gfs.html.

Google (2010c). http://code.google.com/apis/base/.

Google (2010d). Google. http://www.google.com/apps/.

Heroku (2010). Heroku - instant ruby platform. http://

heroku.com/.

Hines, M. (2004). Morgan stanley, ibm ink utility

computing deal. http://news.cnet.com/2100-7339-

5200970.html.

Hofstader, J. (2007). Communications as a service. http://

msdn.microsoft.com/en-us/library/bb896003. aspx.

Hoover, J. N. (2009). NIST team deeply studying cloud

computing – cloud computing – InformationWeek.

http://www.informationweek.com/news/government/

enterprise-architecture/showArticle.jhtml?articleID=

217701603.

IDC (2010). IDC: cloud computing implementations to

double by 2012 - network computing. http://www.

networkcomputing.com/ other/ idc-cloud-computing-

implementations-to-double-by-2012.php?type=article

Identity, P. (2010). PingFederate. http://

www.pingidentity.com/.

Itensil (2010). Itensil. http://itensil.com/.

Jaekel, M. and Luhn, A. (2009). Cloud computing - busi-

ness models, value creation dynamics and advantages

for customers. Technical report, Siemens AG.

Johnston, M. (1999). Preboot execution environment

(PXE). www.pix.net/software/pxeboot/archive/

pxespec.pdf.

Johnston, S. (2010). Taxonomy: The 6 layer cloud comput-

ing stack. http://samj.net/2008/09/taxonomy-6-layer-

cloud-computing-stack.html.

Laird, P. (2009). Laird OnDemand: cloud computing

taxonomy at interop las vegas, may 2009. http://

peterlaird.blogspot.com/2009/05/cloud-computing-

taxonomy-at-interop-las.html.

Linthicum, D. (2009). Deﬁning the cloud computing frame-

work. http://cloudcomputing.sys-con.com/node/8115

19.

Marks, E. A. and Lozano, B. (2010). Executive’s Guide to

Cloud Computing. John Wiley & Sons.

Mell, P. and Grance, T. (2009). The nist deﬁnition of cloud

computing. Technical report, NIST.

Microsoft (2010a). Microsoft connected services frame-

work. http://msdn.microsoft.com/en-us/library/

aa306083.aspx.

Microsoft (2010b). Windows live skydrive. http://

www.windowslive.com.br/public/product.aspx/view/5.

Netsuite (2010). Suitecloud. http://www.netsuite.com.

Nurmi, D., Wolski, R., Grzegorczyk, C., Obertelli, G., So-

man, S., Youseff, L., and Zagorodnov, D. (2009). The

Eucalyptus open-source cloud-computing system. In

Proc. of the 2009 9th IEEE/ACM International Sym-

posium on Cluster Computing and the Grid, CC-

GRID’09, pages 124–131, Washington, DC, USA.

IEEE Computer Society.

Olston, C., Reed, B., Srivastava, U., Kumar, R., and

Tomkins, A. (2008). Pig latin: a not-so-foreign lan-

guage for data processing. In Proc. of the 2008 ACM

SIGMOD international conference on Management of

data, SIGMOD’08, pages 1099–1110, New York, NY,

USA. ACM.

OpSource (2010). http://www.opsource.net/Solutions/Cloud-

Computing.

Oracle (2010). Oracle on demand. http://www.oracle.com/.

OrangeScape (2010). http://www.orangescape.com/.

Prodan, R. and Ostermann, S. (2009). A survey and tax-

onomy of infrastructure as a service and web hosting

cloud providers. 10th IEEE/ACM Internation Confer-

ence on Grid Computing.

Rackspace (2010). Rackspace managed hosting. http://

www.rackspace.com/.

RightScale (2010). Rightscale cloud computing manage-

ment platform. http://www.rightscale.com/.

Rimal, B. P., Choi, E., and Lumb, I. (2009a). A taxon-

omy and survey of cloud computing systems. In Net-

worked Computing and Advanced Information Man-

agement, International Conference on, pages 44–51,

Los Alamitos, CA, USA. IEEE Computer Society.

Rimal, B. P., Choi, E., and Lumb, I. (2009b). A taxonomy

and survey of cloud computing systems. Fifth Inter-

national Joint Conference on INC, IMS and IDC.

rPath (2010). rpath solutions. www.rpath.com/.

SalesForce (2010). Force.com. http://www.salesforce.com/

platform/.

Salesforce (2010a). Salesforce apex. http://wiki.developer

force.com/.

Salesforce (2010b). Salesforce customer relationship man-

agement (crm). http://www.salesforce.com/crm/.

SOASTA (2010). Soasta cloudtest. http://www.soasta.com/.

StorSimple (2010). http://www.storsimple.com.

Trackvia (2010). Trackvia cloud database. http://

www.trackvia.com/products/cloud-database.

UBOOT (2010). Das U–Boot: The universal boot loader.

http://www.denx.de/wiki/U-Boot/.

Velte, A. T., Velte, T. J., and Elsenpeter, R. (2009). Cloud

Computing - A Practical Approach. McGraw-Hill.

Willis, J. M. (2009). Uniﬁed ontology of cloud

computing | IT management and cloud blog.

http://www.johnmwillis.com/cloud-computing/

uniﬁed-ontology-of-cloud-computing/.

Youseff, L., Butrico, M., and da Silva, D. (2008). Toward a

uniﬁed ontology of cloud computing. Grid Computing

Environments Workshop, GCE’08, pages 1–10.

Zoho (2010). Zoho.com. http://www.zoho.com/.

A TAXONOMY MODEL FOR CLOUD COMPUTING SERVICES