An overview of OpenStack architecture: a

message queuing services node

Stanley Lima, Álvaro Rocha & Licinio

Roque

Cluster Computing
The Journal of Networks, Software Tools
and Applications

ISSN 1386-7857

Cluster Comput
DOI 10.1007/s10586-017-1034-x

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Science+Business Media, LLC. This e-offprint
is for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at [Link]”.

1 23
 Author's personal copy
Cluster Comput
DOI 10.1007/s10586-017-1034-x

An overview of OpenStack architecture: a message queuing

services node
Stanley Lima1,2 · Álvaro Rocha2 · Licinio Roque2

Received: 31 March 2017 / Revised: 28 June 2017 / Accepted: 3 July 2017

© Springer Science+Business Media, LLC 2017

Abstract This article provides an overview of the different 1 Introduction

nodes that accommodate the architecture of the OpenStack
project. Throughout this article, we use a statistical approach The National Institute of Standards and Technology (NIST)
to show how OpenStack is currently being developed under describes cloud computing as a model for enabling ubiqui-
the control of the OpenStack Foundation and how this new tous, convenient, on-demand network access to a shared pool
platform is influencing the global market for cloud comput- of configurable computing resources, which can be rapidly
ing. We also discuss community participation in this project provisioned and released with minimal management effort
and present an overview of how these different components or service provider interaction [25].
achieve independence from each other, while using asyn- Cloud computing has changed the information technol-
chronous queues to send messages among themselves in ogy (IT) industry [13], making software even more attractive
order to reliably maintain communication. As a result, we through new technologies, such as full virtualization, or
have achieved a systematic review covering the period from lighter alternatives, such as containerization.
the start of the OpenStack project until the present. This Nowadays, there is a broad discussion with industrial and
longitudinal study can be used to examine variations and academic circles about the advantages and disadvantages
changes in characteristics and different historical versions of of various commercial and public cloud technologies (e.g.,
the OpenStack project (Kilo, Juno, Icehouse etc.) throughout IBM, Amazon EC2, Google App Engine, Windows Azure,
the years. Hadoop, Eucalyptus, Open Nebula and Nimbus) [10,11,59].
However, these commercial clouds usually charge a high
Keywords OpenStack · OpenStack statistics · Open-source price based on the features used. In some specific situations,
software · Cloud computing infrastructure · Scheduling it is more advantageous for a company build their own cloud
cloud resources infrastructure. On the other hand, OpenStack has emerged
as an efficient and experienced infrastructure as a service
(IaaS) with a growing set of free software solutions, either at
a mature stage or under development,in addition to being one
of the cloud stacks that is currently most used. OpenStack is
B Stanley Lima highly configurable to meet different needs. Figure 1 shows
stanleylima@[Link]
the relationship between the services of a typical and Open-
Álvaro Rocha Stack.
amrocha@[Link]
A number of computing researchers and practitioners has
Licinio Roque attempted to define OpenStack in various ways. Here are
lir@[Link]
some definitions:
1 Department of Computer Science, Technische Universität Kavanagh et al. [17] stress that OpenStack has a modular
Dresden, 01069 Dresden, Germany architecture, which is growing with each release, through the
2 Department of Informatics Engineering, University of addition of new components providing extra services in the
Coimbra, 3030-290 Coimbra, Portugal IaaS layer.

123
 Author's personal copy
Cluster Comput

Fig. 1 Conceptual architecture, adapted from Kavanagh [17] 2015

Sahasrabudhe et al. [51] define OpenStack as a cloud and technological advances, with literature reviews forming
computing platform, which provides an IaaS and various a vital part of the research process, as it is essential not only
resources, such as compute, storage and network resources. to define the problem, but also to obtain a precise idea about
It is also a manager of multiple hypervisors and a provider of the current state of knowledge on a given subject, its gaps
various tools for managing and orchestrating cloud resources. and possible research contributions.
Kurup et al. [22] define OpenStack as a collection of open- This work is a continuation of the chapter published in the
source software projects that enterprises or cloud providers WorldCIST’17 [24]. The remainder of this paper is organized
can use to set up and run their cloud compute and storage as follows. First, in Sect. 2, the scientific research method-
infrastructure. The goal of the project is to build an open- ology is presented, while Sect. 3 sets out the results of a
source community with independent researchers and private bibliographic survey referring to each specified database on
companies, so that they are able to share a common goal of scientific methodology. Then, in Sect. 4, the object of study
creating a cloud that is simple to deploy, highly scalable and will be presented in a detailed, updated and statistically man-
full of rich features. ner in order to describe and summarize the pilot project
Voras et al. [58] define OpenStack as a collaborative soft- launched by NASA and Rackspace in July 2010. Lastly, in
ware project aimed at providing a ubiquitous open-source Sect. 5, conclusions related to the issues initially raised by
cloud computing platform that will meet the needs of public the research will be presented.
and private clouds, regardless of size, as well as being simple
to implement and highly scalable.
Khan et al. [18] define OpenStack as open-source cloud 2 Methodology
computing middleware developed by one of the biggest cloud
computing development communities. Currently, OpenStack 2.1 Search strategy
supports platform-specific signatures and tokens for user
authentication. Using the PRISMA guidelines [26], a systematic search of
IT is an area in constant evolution [3]. With OpenStack, the existing literature was performed via the ACM, Springer,
it is essential that new components, reflecting changes from IEEE Xplore and Google Scholar databases covering the
the previous iterations, are combined with new additional period 2010–2016. 2010 was chosen to start the bibliograph-
changes, so that the evolution of the project itself constantly ical survey because it was the year that OpenStack was
changes. Meanwhile, IT developments are linked to scientific launched as a joint project involving NASA and Rackspace

123
 Author's personal copy
Cluster Comput

Table 1 Papers grouped according to the result groups

Results Papers % of total

Tools [4,17,18,27,28]
[9,47,57] 4,2
Qualitative technique or [1,22,23,51,59]
Descriptive model [8,14,49,52,62]
[3,56] 6,3
Empirical model [2,48,55,58,60]
[12,53,64] 4,2
Case studies [5,7,16,21,61] 2,2
Fig. 2 Selection process. The sets a1, a2, a3, b1, b2 and b3 are the sets
of the keywords used in the literature survey Specific solution [15,20,46,50,63]
[19] 3,2

Hosting [17]. Two groups of keywords were listed: Group

3 Results
(a) and Group (b):
As we can see in the research questions in Sect. 2, it was
(a1. OpenStack), (a2. OpenStack statistics), (a3. Open- appropriate to examine what kinds of results were reported
source software) in the selected works. A search on Google Scholar resulted
(b1. Cloud computing infrastructures), (b2. Cloud resource in many duplicate articles because it also indexes the ACM,
management service), (b3. Private and public clouds) Springer and IEEE databases. However, it has a simple search
engine and is extremely fast; in our case, it returned 17,200
articles on the topic “OpenStack” in 0.03 s.
The keywords from Group (a1) were combined with the key-
All the database that were used support Boolean and prox-
words from Group (b1). The results for every database and
imity operators as truncation operators (*, +, or, AND).
each combination of keywords are presented in Fig. 2 in the
However, for the bibliographic research, we only used the
following section.
resulting technology to retrieve documents with variations in
words based on the keywords entered. For example, a search
2.2 Inclusion and exclusion criteria on “Private and public clouds” retrieved documents with the
words “Public cloud”, “Private cloud” or “Private and pub-
All titles, abstracts and conclusions were read to identify lic clouds”. As we can see in Fig. 2, the combinations with
relevant papers. To be included in this systematic review, the Group a2 (“OpenStack statistics”) had no correspondence
articles were evaluated according to PRISMA’s established with any article. This is because the keywords in this group
guidelines for systematic reviews and meta- PRISMA [26] are very specific. Thus, we were able to note that statistics on
and on criteria based on a qualitative assessment, which was OpenStack have not yet been explored in academic papers.
primarily focused on the technical accuracy of the content As can be seen in Table 1, most of the works fell under
presented and on an assessment of factors derived from the the category of “Qualitative technique or descriptive model”.
following research questions: This meant that these works contained contributions, new
methods that corresponded to an updated version of a specific
model, relevant considerations by authors of new updates,
– RQ 1: Are the problems, definitions and motivations of and acquired experiences, which represent the golden rule
the study clearly presented? for new scientific research. However, they were not suffi-
– RQ 2: What is the role of OpenStack in the presented ciently systematic or generic enough to be applied to different
research? domains. By searching out these weeks, we were able to
– RQ 3: Was the research method presented? If yes, can it delimit the research problem, seek new lines of research by
be replicated? avoiding fruitless approaches and identify recommendations
– RQ 4: How did the authors make the link between aca- for future research in the fields related to OpenStack.
demic research and industry?
4 OpenStack
In case of doubts in the content of the title, abstract or conclu-
sions, the full text version of the paper was evaluated against On proprietary software projects, companies contribute with
the inclusion criteria. a 100% of the code [23]. On open-source software projects,

123
 Author's personal copy
Cluster Comput

Fig. 3 Contribution by company

Fig. 4 OpenStack demographics

it is uncommon for a single company to control something

close to 100% of the source code, as a company dominating a bers [46] from 76 countries. Figure 4 presents an analysis of
project is a sign of an unproductive open-source community. the most active countries.
The OpenStack community is extremely active and main-
4.1 Statistics concerning how OpenStack was built tains several options for information about the project, such
as:
In 2010, Rackspace and NASA jointly launched the first
OpenStack distribution, which was titled Austin [19,21]. The • OpenStack maintains a system of meetings involving
goal was to create an open-source cloud platform that could members of the community via Internet relay chat
provide computing services for private and public clouds. (IRC) channel at [Link]
The initial release from Rackspace and NASA only included This provides as a means of communication in real time
two projects: Swift, for a management object storage system, between several OpenStack project areas with approx-
and Nova, for managing compute resource pools. imately 82 channels. Some meetings are not held on
One way to assess the participation of companies in Open- IRC and are listed at [Link]
Stack is to look for corporate contributions in all major Meetings.
components of the project. Figure 3 shows the share of com- • OpenStack Docs at [Link] con-
panies that have mostly contributed to OpenStack throughout tains the official documentation for each project, as well
the years. as a range of forms and training programmes, such as
It is not just about identifying the biggest contributor, and OpenStack IaaS Expert, AWcloud OpenStack and Fire-
whether or not their contribution involved the investment of fly’s LaunchPad.
money, time or human resources, for example, in a partic- • OpenStack Community at [Link]
ular environment or to an OpenStack product. That said, community/ comprises a group of people who share a
relevant questions about potential partners are prompted, common interest in the OpenStack project. It currently
e.g.: (1) On which project area have companies focused? comprises over 30,000 members from over 170 coun-
(2) Which OpenStack partner has the best expertise to sup- tries.
port production needs? (3) Is the investment of this partner • Mailing lists for OpenStack ([Link].
in an OpenStack release compatible with the products and org) at [Link]
services that they are selling? (4) Could this provider be a comprise detailed lists with different directions and tar-
long-term strategic partner to help meet future needs? These get audiences. Every participant can post up to 100 help
are questions that should be answered by analysing a partic- request messages a day.
ular company’s participation in the OpenStack community. • OpenStack Blog at [Link] provides
ever-changing information discussing since the techni-
4.2 OpenStack community cal committee updates to events and weekly community
newsletters.
In less than 5 years, OpenStack has become the largest project • OpenStack-announce at [Link]
in open-source cloud computing, which currently gathers bin/mailman/listinfo/openstack-announce is a list of
together some of the biggest IT companies in the world [48] important announcements about OpenStack’s new
across 180 companies and more than 6000 individual mem- releases, stable releases and security advice.

123
 Author's personal copy
Cluster Comput

Fig. 5 OpenStack evolution

• OpenStack GitHub at [Link] git log -pretty=“commit_id:’%H%ndate:

hosts all the code developed by the OpenStack project and %at%nauthor:%an%nauthor_email:
is available in the public repository from the Rackspace %ae%nsubject:%s%nmessage:%b%n”’-short
website at [Link] stat M -no-merges As a result, we gained the follow-
• OpenStack Event Listing at [Link] ing output:
community/events/ displays all past, current and upcom- commit_id: 123abcxpto date: 11042016
ing events on OpenStack. author: Karine kaliane author_email:
• OpenStack Ambassadors at [Link] [Link]@[Link] subject: add
org/about is used to share knowledge, experiences and object-based import in the code
collaborative resolutions of problems among community message: change class 12hd432 Change-
members. The topics range from the presentations of case Id: Idc87sd76367238328apr2016 diff_stat:
studies to broader issues, such as management applica- seven files changed, 290 insertions
tions. (+), 307 deletions (-).
Regarding the evolution of the OpenStack development
4.3 OpenStack architecture and components community in general, in this section, we present statistical
graphs that tell a story of growth throughout the release cycle
The architecture is still modular [56], but, as shown in Fig. 5, of the OpenStack project.
it has grown with every release by adding new components In Fig. 5, we can see the evolution of the community from
that provide services in the IaaS layer. Nowadays, OpenStack 2010 to 2016. According to the chart, on average, 10,000
is organized around several projects of varying size and dif- members joined between 2010 and 2013. This increased to
ferent levels of activity [14,61]. This facility allows us to 30,000 in 2015, while there was an increase of 7.1% on aver-
look for collaborative OpenStack contributions. One basic age in 2015–2016: (formula (present) = (passed)∗ (1 + growth
way to identify collaborative contributions is to analyse the rate)n , where n = number of time periods).
contributions aggregated from all OpenStack projects. This significant jump was due to the participation of new
Statistics are fundamental to transforming data into infor- companies in the OpenStack project [12]. If we look at the
mation. Data on OpenStack were supplied by Stackalytics contribution of each company to the latest release of Open-
[54]. Stackalytics processes three major metrics for Open- Stack, Mitaka, by the number of members, and express it
Stack contributions: the number of modified files, the number as a percentage of the total contribution, we can accept the
of members and the number of modified lines. The related following graph as valid.
statistics were retrieved from ([Link]) using the fol- Mitaka, the new OpenStack release, shows a continu-
lowing command. ous increase in activity compared to Liberty. From Icehouse

123
 Author's personal copy
Cluster Comput

to Juno, there was an increase of 7.23% in the number of Table 2 OpenStack releases and their corresponding software versions
members. From Juno to Kilo, there was a greater leap in OpenStack release name
terms of membership (9.23%), but the biggest difference fol- Releases Date Components
lowed the release period from Austin to Diablo and Essex,
Austin (Texas) 2010.10 [40,43]
which saw a 32.7% increase in members. This progres-
Bexar (Texas, US) 2011.01 [33,40,43]
sive increase is directly proportional to the incorporation of
new companies into the OpenStack project. Red Hat was Cactus (Texas, US) 2011.04 [33,40,43]
the first contributor, with about a quarter of all members, Diablo (Santa Clara, US) 2011.09 [33,40,43]
closely followed by Rackspace, HP and Miratins, with IBM Essex (Boston, US) 2012.04 [33,35,37,40,43]
at some distance behind. Other contributors comprise more Folsom (San Francisco, US) 2012.09 [33,35,37,40,43]
than 550 companies and more than 30,000 members in more [31,39]
than 170 countries collaborating on the OpenStack project Grizzly (San Diego, US) 2013.04 [33,35,37,40,43]
[28,52]. Looking at these numbers, it is clear that the Open- [31,39]
Stack project is now more levelled, with less dominance by Havana (Portland, US) 2013.10 [33,35,37,40,43]
Rackspace. There is also a clearly emerging picture of the [30,31,34,39]
participation of other companies, which appear to be heavily Icehouse (Hong Kong, CN) 2014.04 [33,35,37,40,43]
invested in improving the OpenStack codebase. [30,31,34,39,44]
Currently, OpenStack is in its 13th release, known as Juno (Atlanta, US) 2014.10 [33,35,37,40,43]
Mitaka (released on 07.04.2016), while Newton was sched- [30,31,34,39,44]
uled for release on 06.10.2016 [6]. Just as SOA has features [41]
such as low coupling and reusability, any release by the Open- Kilo (Paris, FR) 2015.04 [33,35,37,40,43]
Stack project must meet the following requirements [53]: [30,31,34,39,44]
[36,41]
Open source Liberty (Vancouver, CA) 2015.10 [33,35,37,40,43]
[30,31,34,39,44]
Dykstra and Sherman [9] As the entire code of OpenStack is
[32,36,38,41,45]
open, it is regarded as an IaaS environment, which seeks to
[29,42]
provide a ubiquitous computing platform in an open-source
Mitaka (Tokyo, JP) 2016.04 [33,35,37,40,43]
cloud for private and public use. The development distri-
[30,31,34,39,44]
bution and adoption of OpenStack are supervised by the
[32,36,38,41,45]
OpenStack Foundation, an independent agency that provides
[29,42]
shared resources.
Newton (Austin, US) 2016.10 Under development
Modularity and flexibility Ocata (Barcelona, ES) 2017.01 To be determined

Oechsner and Ripke [27] OpenStack provides an IaaS solu- being able to use each other. It is also possible to modify,
tion through a set of unified services, with each service adapt or customize any service, provided that the basic struc-
offering an API that facilitates this integration, as well as ture of the APIs is maintained, so that each of the constituent
supporting several virtualization standards, including KVM, services work together to provide an integrated IaaS solu-
UML, XEN, Hyper-V and QEMU [57,64]. tion. Figure 6 shows the relationship between the OpenStack
services.
Scalable
Compute (Nova) [40] is the cloud computing infrastructure
Povedano-Molina et al. [47] In many ways, OpenStack archi- manager, which can carry out all the activities necessary to
tecture itself is designed to be horizontally scalable, as it maintain the life cycle of the instances controlled by Nova.
needs to provide intelligent storage solutions to meet grow- It is also able to manage all necessary resources, such as
ing business needs. cloud scalability and the network. The Nova project is a
The OpenStack releases possess different code names, management platform, although it does not itself possess
which are proposed and chosen by the technical committee. any virtualization capability; rather, it uses the libvirt APIs
Each release comprises components/services that are inter- for supporting a wide variety of virtualization technologies,
related. This modularity enables the deployment of specific including KVM, Xen, LXC and VMware.
services in each case. Table 2 shows the progression of the Image Service (Glance) [33] lists, records and retrieves vir-
components over recent years, with each of these services tual machine images. It manages glance images in a cluster of

123
 Author's personal copy
Cluster Comput

Fig. 6 OpenStack logical architecture

123
 Author's personal copy
Cluster Comput

OpenStack; however, is not responsible for the actual storage Database (Trove) [44] is database as a service for Open-
as it provides an abstraction for specific storage technologies. Stack. Trove lays the groundwork for automating the pro-
Object Storage (Swift) [43] is an OpenStack object store vision and management of leading features in a relational
project, known as Swift, which provides cloud storage soft- or non-relational database without the burden of handling
ware, so that applications can store and retrieve lots of data complex administrative tasks. Initially, the Trove project is
with a simple API. Such objects are stored in different units, focusing on resource isolation at high performance, while
with Swift ensuring that data are replicated in a clustered automating complex administrative tasks, including patch-
server. Initially, Swift distributes copies of each object by ing, backups, restores, configuration and deployment.
region (then by zone, server and across drives).
Elastic Map Reduce (Sahara) [41] aims to provide users
Dashboard (Horizon) [35] provides a web-based user inter- with a simple means by which to provide data processing,
face to OpenStack services, including Glance, Nova, Neutron such as Spark or Hadoop, on OpenStack. This is accom-
and Sahara. It brings together all OpenStack projects in a sin- plished by specifying configuration parameters, such as the
gle pane based on a Django module called django-openstack. framework version, cluster topology and node hardware.
Identity Service (Keystone) [37] is an OpenStack service Bare Metal Provisioning (Ironic) [36] is an OpenStack ser-
that manages user databases and authentication types, such vice, which provides capability to the bare metal servers
as (authn), and high-level authorizations, such as Microsoft instead of virtual machines, forked from the Nova bare metal
AuthZ API. It integrates with existing back end directory [Link] also supports vendor-specific plug-ins, which
services, such as LDAP, and has recently been re-structured are able to implement additional functionality. For a gen-
to allow for expansion in order to support proxy services and eral overview, please see “Introduction to Ironic”, in which
external AuthZ/AuthN mechanisms, such as oAuth, SAML its drivers support a variety of hardware vendors, such as
and openID. iBOOT, DRAC and SMTP.
Networking (Neutron, Formerly Quantum) [39] provides
Multiple Tenant Cloud Messaging (Zaqar) [45] is an
networking as a service to other OpenStack components
OpenStack messaging and notification service. It is simi-
(such as Nova). It does this by creating and attaching the
lar to Amazon’s SQS product with additional semantics to
virtual switch port to the vNIC of the VM. Neutron also
support event transmission. For example, when installing
includes a set of API interfaces and plug-ins, which enable
python-zaqarclient, the application will have access to the
interoperability and network device orchestration in IaaS
python-openstackclient plug-in order to create single-row
environments.
operations. Thus, users can customize Zaqar to achieve a
Block Storage (Cinder) [31] implements services and wide range of performance, durability, availability and more
libraries to provide on-demand, self-service access to block efficient targets.
storage resources. Software-defined block storage is facil-
Shared File System Service (Manila) [38] is a community-
itated via abstraction and automation on top of various
driven project, which offers the management of file shares,
traditional back end block storage devices. In the most com-
such as NFS, as a core service to OpenStack.
mon scenario, the Cinder volumes provide persistent storage
to guest virtual machines. DNSaaS (Designate) [32] provides DNS as a service for
Orchestration (Heat) [34] is an OpenStack project specif- OpenStack, such as REST API for domain, record man-
ically dedicated to orchestration, which makes it possible agement and integration with Keystone for authentication,
to manage multiple instances in an automatic way. In turn among other services. It is equivalent to Amazon Route 53
the heat engine component provides the main orchestration and authenticated via the Keystone Token, while Designate
functionality and supports an orchestration engine to cre- processes remote procedure call (RPC) requests through MQ.
ate multiple composite cloud-based model applications in An API is used for the DNS back end, such as MySQLBind
the form of text files, which can be treated as code through and PowerDNS.
the native OpenStack RESTful API function, a heat-api-cfn Search Service (Searchlight) [42] aims to improve search
component similar to the AWS CloudFormation compliant capabilities and the performance of various OpenStack cloud
API Query. services. It accomplishes this by offloading user search
Telemetry (Ceilometer) [30] is a telemetry project that cur- queries from existing API servers and indexing their data
rently provides a set of functionalities split across multiple in ElasticSearch, which is a search server based on Apache
projects, in which each project is designed to provide a dis- Lucene. The goal here is to perform highly effective and
crete service in the telemetry space by publishing collected flexible queries, combined with p-indexing in real time.
data for various targets, including data stores and message Security API (Barbican) [29] is a REST API OpenStack
queues, among other functions. project designed for the safe storage, provision and manage-

123
 Author's personal copy
Cluster Comput

Fig. 7 Queuing services node using RabbitMQ

ment of secret information, such as passwords, encryption tion and queue-to-queue encryption [18] is recommended.
keys and SSL certificates. For authentication, data security is available via OpenStack
Let us dig deeper into the topic. The basic OpenStack Queue Access Control, which uses a simple authentication
architecture consists of three independent nodes: a com- and security layer framework via RabbitMQ.
pute node to create and manage the hypervisors; a network In OpenStack, we have verified that the queue system,
node; and a controller or message queuing services node. which connects the components, is exploited in real time,
Each component works independently of each other, while allowing for constant communication between the different
and the synchronization between the components central- nodes, so that the real-time processing of the implemented
izes them. Even if we create the components in a single structure makes the components scalable. As for the com-
node, they behave as independent components. The com- munities and companies collaborating in the development
munication between the components occurs as a result of of the project, we have noticed that they work independently
using the controller node, which provides a queuing and and with different goals. This is possible because the compo-
internal management service using REST-call APIs through nents of OpenStack are linked together through shared APIs,
RabbitMQ. The Nova, Cinder and Quantum nodes commu- hosted and managed by GitHub, which guarantees that any
nicate via the Advanced Message Queue Protocol (AMQP), change to any part of the OpenStack code is subject to the
while nova-compute, nova-scheduler and nova-api commu- same rigorous testing processes, be it a change to the code
nicate using RPCs. These RPCs queue up in OpenStack, or an adjustment to the documentation.
via RabbitMQ, by using PUBLISHERS and CONSUMERS
providers. The idea is for the PUBLISHERS to not send a
message directly to the queue. For this the framework, Rab- 5 Conclusion
bitMQ implements an EXCHANGE to facilitate the routing
via a key (e.g., msg_id), which is passed by the PUB- OpenStack is a global collaboration project between indepen-
LISHERS responsible for the message. In Fig. 7, we can dent developers and private companies, which aims to offer
see this controller node interacting with the other compo- an open-source cloud platform in order to provide computing
nents. services for private and public clients. In turn, it facilitates
As this queuing service does not provide an authoriza- the creation and management of different sets of computing
tion checking service in order to facilitate tamper and spy resources, storage and networks via a dashboard involving
protection on the messaging server, performing authentica- an interface web.

123
 Author's personal copy
Cluster Comput

This article offers a careful analysis of the state of the art of References
the different components of the OpenStack project. In more
detail, we discussed the main activities and the influence of 1. Baset, S.A.: Open source cloud technologies. In: Proceedings of
the Third ACM Symposium on Cloud Computing, p. 28. ACM
the existing communities, such as large companies in the
(2012)
IT sector, which support the OpenStack Foundation (e.g., 2. Beernaert, L., Matos, M., Vilaça, R’., Oliveira, R.: Automatic elas-
Rackspace, Red Hat and IBM). To contextualize the study, ticity in openstack. In: Proceedings of the Workshop on Secure and
we presented a systematic review of the literature, as well as Dependable Middleware for Cloud Monitoring and Management,
p. 2. ACM (2012)
conducted statistical studies to analyse the participation of
3. Bennett, K.H., Rajlich, V.T.: Software maintenance and evolution:
the leading companies involved in the project so far. We have a roadmap. In: Proceedings of the Conference on the Future of
noted that some of the OpenStack members are providers Software Engineering, pp. 73–87. ACM (2000)
of infrastructure or components of the cloud itself, while 4. Bruneo, D., Longo, F., Merlino, G., Peditto, N., Romeo, C., Ver-
boso, F., Puliafito, A.: Enabling collaborative development in an
others are relying on the different projects that make up the
openstack testbed: the cloudwave use case. In:Proceedings of
OpenStack project to create specific products, which they sell the Seventh International Workshop on Principles of Engineer-
onto other companies. ing Service-Oriented and Cloud Systems, pp. 24–30. IEEE Press
While Dell has focused on providing hardware and con- (2015)
5. Calero, J.M., Aguado, J.G.: Monpaas: an adaptive monitoring plat-
sulting services in the implementation of OpenStack compo- formas a service for cloud computing infrastructures and services.
nents, IBM has offered its expertise in enterprise software IEEE Trans. Serv. Comput. 8, 65–78 (2015)
to meet the desired quality expectations of its customers 6. Cash, S., Jain, V., Jiang, L., Karve, A., Kidambi, J., Lyons, M.,
in implementing a cloud. There is currently an IBM drive Mathews, T., Mullen, S., Mulsow, M., Patel, N.: Managed infras-
tructure with ibm cloud openstack services. IBM J. Res. Dev.
linked to the newly acquired SoftLayer, which provides an 60(2–3), 1–6 (2016)
IBM cloud computing suite of services directed at Open- 7. Chang, H.-T., Wang, S.-Y.: Using sdn technology to mitigate con-
Stack. Moreover, Canonical gained prominence through its gestion in the openstack data center network. In: Communications
Ubuntu OS, which in turn has become the main operat- (ICC), 2015 IEEE International Conference on, pp. 401–406. IEEE
(2015)
ing system for OpenStack project implementations, whereas 8. Coutinho, E.F., de Carvalho Sousa, F.R., Rego, P.A.L., Gomes,
VMware has become one of the most experienced players D.G., de Souza, J.N.: Elasticity in cloud computing: a survey. Ann.
in virtualization. From a macroscopic point of view, we can Telecommun. 70, 289–309 (2015)
see that the OpenStack project is an open-source alterna- 9. Dykstra, J., Sherman, A.T.: Design and implementation of frost:
Digital forensic tools for the openstack cloud computing platform.
tive to VMware products. However, VMware is focused on Digital Investig. 10, S87–S95 (2013)
OpenStack cloud management tools, such as the ESX hyper- 10. Eisa, M., Younas, M., Basu, K., Zhu, H.: Trends and directions
visor. Next, we described in detail how the different nodes in cloud service selection. In: 2016 IEEE Symposium on Service-
of this architecture interact with each other through the mes- Oriented System Engineering (SOSE), pp. 423–432. IEEE (2016)
11. Galante, G., De Bona, L.C.E., Mury, A.R., Schulze, B., da Rosa
saging services node. In the process, we were able to see Righi, R.: An analysis of public clouds elasticity in the execution
that message queuing enables interprocess communication of scientific applications: a survey. J. Grid Comput. 14(2), 193–216
in OpenStack through asynchronous requests via the AMQP, (2016)
in which services that are accessible through the queue vali- 12. Gonzalez-Barahona, J.M., Izquierdo-Cortazar, D., Maffulli, S.,
Robles, G.: Understanding how companies interact with free soft-
date message tokens dispense with other access authorization ware communities. IEEE Softw. 5, 38–45 (2013)
checks. In turn, we now have a better understanding of the 13. Hashem, I.A.T., Yaqoob, I., Anuar, N.B., Mokhtar, S., Gani, A.,
differences between the pluggable components that make up Khan, S.U.: The rise of big data on cloud computing: review and
this architecture. open research issues. Inf. Syst. 47, 98–115 (2015)
14. Jones, M.T.: Cloud computing and storage with openstack: Dis-
Meanwhile, in the future, we would like to examine com- cover the benefits of using the open source openstack iaas cloud
plexity with regard to some of the top cloud providers, platform. Developer Works (2012)
such as VMware, Platform9, CloudStack and Microsoft, 15. Ju, X., Soares, L., Shin, K.G., Ryu, K.D., Da Silva, D.: On fault
as well as understand how these players are differen- resilience of openstack. In: Proceedings of the 4th annual Sympo-
sium on Cloud Computing, p. 2. ACM (2013)
tiating or adapting their OpenStack project products in 16. Karande, P., Gaherwar, S., Kurhekar, M: Physical to virtual migra-
response to reliability, performance, scalability and support tion of ubuntu system on openstack cloud. In: Proceedings of
issues. the Third International Symposium on Women in Computing and
Informatics, pp. 510–515. ACM (2015)
17. Kavanagh, A.: Openstack as the api framework for nfv: the benefits,
Acknowledgements This work was supported by research grants from and the extension needed. Ericsson Rev. 2, 15 (2015)
the Ciências sem Fronteiras (CsF) and Agência Espacial Brasileira 18. Khan, R.H., Ylitalo, J., Ahmed, A.S.: Openid authentication as
(AEB) programmes. We also appreciate the financial support from a service in openstack. In: Information Assurance and Security
the Iberian Association for Information Systems and Technologies (IAS), 2011 7th International Conference on, pp. 372–377. IEEE
(AISTI). (2011)

123
 Author's personal copy
Cluster Comput

19. Kim, H., Chun, K., Kim, H., Chung, Y.: Utilization of workflow 48. Qevani, E., Panagopoulou, M., Stampoltas, C., Tsitsipas, A., Kyr-
management system for virtual machine instance management iazis, D., Themistocleous, M.: What can openstack adopt from a
on cloud. Concurrency Comput.: Pract. Exp. 27(17), 5350–5373 ganeti-based open-source iaas? In: 2014 IEEE 7th International
(2015) Conference on Cloud Computing (CLOUD), pp. 833–840. IEEE
20. Kumar, R., Gupta, N., Charu, S., Jain, K., Jangir, S.K.: Open source (2014)
solution for cloud computing platform using openstack. Int. J. Com- 49. Ristov, S., Gusev, M., Donevski, A.: Openstack cloud security
put. Sci. Mob. Comput. 3(5), 89–98 (2014) vulnerabilities from inside and outside, pp. 101–107. Cloud Com-
21. Kumar, R., Parashar, B.B.: Dynamic resource allocation and man- puting (2013)
agement using openstack. Nova 1, 21 (2010) 50. Ristov, S., Gusev, M., Donevski, A.: Security vulnerability assess-
22. Kurup, L.D., Chandawalla, C., Parekh, Z., Sampat, K.: Compara- ment of openstack cloud. In: Computational Intelligence, Com-
tive study of eucalyptus, open stack and nimbus. Int. J. Soft Comput. munication Systems and Networks (CICSyN), 2014 Sixth Interna-
Eng. 4(6), 23–27 (2015) tional Conference on, pp. 95–100. IEEE (2014)
23. Lakhani, K.R., Von Hippel, E.: How open source software works: 51. Sahasrabudhe, S.S., Sonawani, S.S.: Comparing openstack and
free user-to-user assistance. Res. Policy 32(6), 923–943 (2003) vmware. In: Advances in Electronics, Computers and Communica-
24. Lima, S., Rocha, A.: A view of openstack: toward an open-source tions (ICAECC), 2014 International Conference on, pp. 1–4. IEEE
solution for cloud. In: Rocha, Á., Correia, A., Adeli, H., Reis, L., (2014)
Costanzo, S. (eds.) Recent Advances in Information Systems and 52. Scharf, M., Stein, M., Voith, T., Hilt, V.:. Network-aware instance
Technologies. WorldCIST 2017. Advances in Intelligent Systems scheduling in openstack. In: Computer Communication and Net-
and Computing, vol. 569, pp. 481–491. Springer, Cham (2017) works (ICCCN), 2015 24th International Conference on, pp. 1–6.
25. Mell, P., Grance, T.: The nist definition of cloud computing (2011) IEEE (2015)
26. Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G.: Preferred report- 53. Singh, P., Singh, V.P., Pachauri, G.: Critical analysis of cloud com-
ing items for systematic reviews and meta-analyses: the prisma puting using openstack. Int. J. Comput. Sci. Mob. Comput. 3(3),
statement. Ann. Int. Med. 151(4), 264–269 (2009) 2014
27. Oechsner, S., Ripke, A: Flexible support of vnf placement functions 54. Stackalytics. [Link] Apr 2016
in openstack. In: Network Softwarization (NetSoft), 2015 1st IEEE 55. Tong, J., Ying, L., Xiaoyong, Y., Hongyan, T., Zhonghai, W.:
Conference on, pp 1–6. IEEE (2015) Characterizing and predicting bug assignment in openstack. In:
28. Omar, I.Y., Laborde, R., Wazan, A.S., Barrere, F., Benzekri, Trustworthy Systems and Their Applications (TSA), 2015 Second
A.: G-cloud on openstack: Adressing access control and regula- International Conference on, pp. 16–23. IEEE (2015)
tion requirements. In: Networks, Computers and Communications 56. Van, V.N., Long, N.Q., Le, D.-N. et al.: Performance analysis of
(ISNCC), 2015 International Symposium on, pp. 1–6. IEEE (2015) network virtualization in cloud computing infrastructures on open-
29. Openstack Barbican. [Link] stack. In: Innovations in Computer Science and Engineering, pp.
Apr 2016 95–103. Springer, New York (2016)
30. Openstack Ceilometer. [Link] 57. Varrette, S., Plugaru, V., Guzek, M., Besseron, X., Bouvry,P.: Hpc
ceilometer. Apr 2016 performance and energy-efficiency of the openstack cloud mid-
31. Openstack Cinder. [Link] Apr dleware. In: Parallel Processing Workshops (ICCPW), 2014 43rd
2016 International Conference on, pp. 419–428. IEEE (2014)
32. Openstack Designate. [Link] 58. Voras, I., Mihaljević, B., Orlić, M., Pletikosa, M., Pavić, T., Zim-
Apr 2016 mer, K., Čavrak, I., Paunović, V., Bosnić, I., Tomić, S., et al.:
33. Openstack Glance. [Link] Apr Evaluating open-source cloud computing solutions. In: MIPRO,
2016 2011 Proceedings of the 34th International Convention, pp. 209–
34. Openstack Heat. [Link] Apr 2016 214. IEEE (2011)
35. Openstack Horizon. [Link] Apr 59. Wen, X., Gu, G., Li, Q., Gao, Y., Zhang, X.: Comparison of open-
2016 source cloud management platforms: Openstack and opennebula.
36. Openstack Ironic. [Link] Apr In: Fuzzy Systems and Knowledge Discovery (FSKD), 2012 9th
2016 International Conference on, pp. 2457–2461. IEEE (2012)
37. Openstack Keystone. [Link] 60. Wuhib, F., Stadler, R., Lindgren, H.: Dynamic resource alloca-
Apr 2016 tion with management objectives implementation for an openstack
38. Openstack Manila. [Link] Apr cloud. In: Network and service management (cnsm), 2012 8th inter-
2016 national conference and 2012 workshop on systems virtualiztion
39. Openstack Neutron. [Link] Apr management (svm), pp. 309–315. IEEE (2012)
2016 61. Xu, Q., Yuan, J.: A study on service performance evaluation
40. Openstack Nova. [Link] Apr 2016 of openstack. In:Broadband and Wireless Computing, Commu-
41. Openstack Sahara. [Link] Apr nication and Applications (BWCCA), 2014 Ninth International
2016 Conference on, pp. 590–593. IEEE (2014)
42. Openstack Searchlight. [Link] 62. Yadav, D.S., Doke, K.: Mobile cloud computing issues and solution
searchlight. Apr 2016 framework (2016)
43. Openstack Swift. [Link] Apr 2016 63. Yamato, Y.: Automatic verification technology of software patches
44. Openstack Trove. [Link] Apr 2016 for user virtual environments on iaas cloud. J. Cloud Comput. 4(1),
45. Openstack Zaqar. [Link] Apr 2016 1–14 (2015)
46. Plasencia, I.C., del Castillo, E., Heinemeyer, S., Lopez-Garcia, A., 64. Younge, A.J., Henschel, R., Brown, J.T., Von Laszewski, G., Qiu, J.,
vd Pahlen, F.:. Phenomenology tools on cloud infrastructures using Fox, G.C.: Analysis of virtualization technologies for high perfor-
openstack. CoRR, vol. abs/1212.4784 (2012) mance computing environments. In: Cloud Computing (CLOUD),
47. Povedano-Molina, J., Lopez-Vega, J.M., Lopez-Soler, J.M., Cor- 2011 IEEE International Conference on, pp. 9–16. IEEE (2011)
radi, A., Foschini, L.: Dargos: a highly adaptable and scalable
monitoring architecture for multi-tenant clouds. Future Gener.
Comput. Syst. 29(8), 2041–2056 (2013)

123
 Author's personal copy
Cluster Comput

Stanley Lima is a PhD stu- Planning and Management, Information Systems Quality, Require-
dent at the department of Infor- ments Engineering, Software Engineering, e-Government, e-Health,
matics engineering, in the doc- and Information Technology in Education. He is President of AISTI
toral program in Information Sci- (Iberian Association for Information Systems and Technologies), Chair
ence and Technology of Univer- of IEEE Portugal Section Systems, Man, and Cybernetics Society Chap-
sity of Coimbra, in the search ter, and Vice-President of ADI-PME (Association for the Development
group Information Systems (IS) and Innovation in the Small and Medium Enterprises). He is also the
and co-supervised by Zurich Editor-in-Chief of JISEM (Journal of Information Systems Engineer-
University of Applied Sciences ing & Management) and the Editor-in-Chief of RISTI (Iberian Journal
- ZHAW School of Engineer- of Information Systems and Technologies). And he is Vice-Chair of
ing. During master degree he Experts in the Horizon 2020 of the European Commission, Expert in
works at the General Headquar- the COST - intergovernmental framework for European Cooperation
ters of the Brazilian Army in in Science and Technology, Expert in the Italian Ministry of Educa-
Brasilia, in the CORTEX project tion, University and Research, and General-Chair of WorldCIST (World
(middleware for communication Conference on Information Systems and Technologies).
between armoured). After completing his masters in Software Engi-
neering, Stanley worked in the department of computer science at
Technische Universit?t Dresden, Germany in 2015- working in Service Licinio Roque obtained a PhD
Oriented Architecture related technologies. in Informatics Engineering from
the University of Coimbra while
developing “Context Engineer-
Álvaro Rocha holds Habili- ing”, a Socio–Technical Appro-
tation in Information Science, ach to Information Systems
PhD in Information Systems and Development. He teaches studio
Technologies, MSc in Informa- and project–based postgraduate
tion Management, and Bachelor courses on Human–Computer
in omputer Science. He is Profes- Interaction, Game Design and
sor of Information Systems and Sociotechnical Systems Devel-
Software Engineering at Univer- opment at the University of
sity of Coimbra, researcher at Coimbra Master on Informat-
CISUC (Centre for Informatics ics Engineering and PhD pro-
and Systems of the University gram on Information Sciences
of Coimbra) and a collaborator and Technologies, and was Adjunct Associate Professor at Carnegie
researcher at LIACC (Labora- Mellon University Master Programs on Software Engineering. His
tory of Artificial Intelligence and research interests include Human–Computer Interaction and Digital
Computer Science) and at CIN- Games, focusing on Participatory Media Design.
TESIS (Center for Research in Health Technologies and Informa-
tion Systems). His main research interests are Information Systems

123