1-Virtual Desktop Infrastructure (VDI) Deployment Using OpenNebula As A Private
1-Virtual Desktop Infrastructure (VDI) Deployment Using OpenNebula As A Private
net/publication/339640756
CITATIONS READS
0 76
6 authors, including:
Some of the authors of this publication are also working on these related projects:
Desarrollo e implementación de dos aplicaciones multiplataforma para analizar presuntos problemas de seguridad y privacidad de datos en usuarios de redes sociales.
Caso de Estudio: Universidad Politécnica Salesiana Sede Cuenca View project
Educación vial de peatones y conductores de la ciudad de Cuenca a través de sistemas de realidad virtual. View project
All content following this page was uploaded by Paul Esteban Vintimilla Tapia on 09 July 2020.
1 Introduction
Today, thanks to the constant advances of Information and Communication
Technologies (ICTs), the popularity of Internet-based teaching techniques is
growing, giving way to the construction of electronic learning environments [4].
Thus, tools are born that intervene in the acquisition and management of knowl-
edge to allow the deployment of new educational paradigms, eliminating the
space-time barriers that limit traditional models [5]. However, the growing num-
ber of students generates a large amount of data that attacks the computing
infrastructure of educational institutions, because they do not support adequate
processing. For this reason, and bearing in mind the difficulties involved in in-
vesting in hardware or software, when implementing any e-learning application
or service, efforts should be focused on achieving optimal allocation and con-
sumption of resources according to highly dynamic demand [15]. In addition, it
is essential to ensure an adequate quality of service, since the student’s interest
could be lost and, therefore, the academic objectives set could not be met [17].
Faced with this situation, a promising alternative is emerging in cloud com-
puting, thanks to the fact that it represents a technological trend that allows
resources to be exchanged via the Internet [13]. Different providers offer as a
service the capacity of their equipment in terms of processing, storage, network
and applications, to meet the needs of different organizations, avoiding invest-
ment in new infrastructure [9]. Thus, the benefits of transferring this scenario to
educational institutions can be summarized in 4 points [6]:
1. Reduction and rationalization of expenses: the use of cloud-based services
simplifies a series of complicated processes depending on the maintenance
of the local infrastructure and the deployment of e-learning applications,
reflecting a saving of resources and capital.
2. On-demand service delivery: Cloud computing is a technology model that
supports the aggregation of resources to meet customers’ exponential de-
mand. Thus, at times of peak activity such as evaluations or registration
processes, only the service provider should be asked to momentarily allocate
a greater processing capacity or network, preventing possible congestion or
system crashes.
3. Greener education: by employing third-party hardware, opportunities are
offered to reduce energy consumption, giving way to environmentally friendly
institutions. For its part, from the provider side, each cloud hosts multiple
tenants, so the processing capacity is shared efficiently, indirectly reducing
polluting emissions.
4. Improve academic proficiency: Under the infrastructure-as-a-service model,
cloud computing can provide each student with a virtual desktop, commonly
known as virtual desktop infrastructure (VDI). Teachers can then design
activities that require high-performance software without any limitations.
With respect to the last point, the use of VDIs significantly helps to change
the traditional concept of teaching, since it makes it easier to overcome the
restrictions involved in the use of proprietary software both licensing and equip-
ment (processor, hard disk and RAM) [12]. In few words, a VDI is a cloud
computing service that supports virtualization of desktops, including operating
systems and software, on remote servers accessed via the Internet [11]. Therefore,
it is possible to provide each student with a virtual and personalized environment
that has sufficient features to support the development of the tasks assigned by
the teacher, ensuring ubiquitous access from any device [7]. In addition, from a
practical point of view, infrastructure managers also benefit by reducing equip-
ment setup times, increasing information transfer rates, and mitigating security
and compatibility issues [2].
On base the mentioned, it can be said that there is a close relationship with
the concept of centralized computing, where users are connected to the host
computer through terminals without intelligence that do not have processing or
storage capabilities [1]. For this reason, it is essential to know the architecture
on which the VDIs are executed, which consists of six components [8]:
As can be seen, there is not much complexity when implementing the compo-
nents mentioned, so it is common to find companies focused on cloud computing
and virtualization that offer within their products the deployment of VDIs as
a service. However, licensing costs cannot be ignored, as they could not be in-
cluded in the funds of educational institutions with limited budgets. Therefore,
this article proposes the development of a system of studio environments and
virtual desktops whose architecture is hosted and virtualized in OpenNebula, a
private cloud solution. The main objective is to provide tools that allow stu-
dents to consolidate their knowledge by resolving practices generated in a given
context according to learning interests and needs.
The rest of the article is organized as follows. Section II presents works that
reflect relevant technological proposals aimed at the use of virtual desks to sup-
port education. Section III describes the diagram of the proposed general ar-
chitecture, including different modules and services. In Section IV, the results
obtained in a pilot experiment carried out with students related to the field of
Computer Science are analysed. Finally, the main conclusions and future work
can be found in section V.
2 Related Works
First of all, the authors of [14] analyze the opportunities and benefits of virtual
platforms, which allow students to install, configure, and manipulate remote
desktops, safeguarding the stability and security of the physical infrastructure.
Thus, it is possible to develop laboratory practices that could compromise the
functioning of physical equipment, as a result of the lack of experience in its han-
dling. Hence, emphasis is placed on several important concepts related to virtu-
alization: enabling technologies, security systems, limitations and challenges of
use in educational environments. Principally, a business-oriented virtual security
system is deployed, with the objective of bringing students closer to the real en-
vironment of a company in which security policies are implemented. This process
was carried out through virtual laboratories that were accessed through work-
stations connected to the same network. Each laboratory was equipped with the
necessary tools and resources to implement the required practices without diffi-
culty. The results obtained were favorable because the students gained greater
interest and a more professional work vision.
For its part, in [10] tries to explain how to make an infrastructure to provide
a remote desktop service (Desktop as a Service - DaaS) that is compatible with
any operating system. To do this, unlike traditional solutions, several servers
hosted in a private cloud are occupied instead of hypervisors, which run virtual
machines. In addition, protocols based on virtual networks are specified for end
user access. Meanwhile, for cloud administrators, an environment is implemented
that deploys tools to support DaaS management. The achieved results expose
access facilities to the different desktops, as well as for the handling and storage
of information in each one. However, several shortcomings were reported when
responding to simultaneous requests, determining a fundamental aspect to keep
in mind for future investigations.
Meanwhile, in [16] the authors detail the design of a virtual laboratory ar-
chitecture capable of providing students with a practical learning experience
supported by an online educational offering. In this sense, the design approaches
applied are described, addressing criteria for the selection of an optimal im-
plementation method. Thus, each student accesses a virtual desktop through
his/her personal computer, being able to install and execute different software
solutions, according to the instructions contemplated in the laboratory prac-
tices. The evaluation results indicated that the use of VDIs help to improve the
performance of the learning environment, supporting multiple users.
Finally [3] explores the practical aspects of migrating to virtual architectures,
describing the costs, benefits and challenges of implementing and managing vir-
tual laboratories to support teaching and research in areas such as engineer-
ing, computing and information security. For this reason, the alternatives that
have been deployed in the private clouds of different universities are analyzed,
starting from the description of their infrastructure, application environment,
virtualization software, advantages and limitations, relevant characteristics and
considerations to take into account for future moves to hybrid or public clouds.
In conclusion, the existence of certain risks that involve the movement of a pri-
vate cloud to the outside is highlighted, so it is recommended to migrate the
content in small parts, maintaining adequate administration and performance.
As can be noted, the papers presented describe the use of VDIs as tools
that support the educational process. However, they do not take into consid-
eration the aspects or restrictions represented by the use of licensed software,
which could slow down technological acceptance by educational institutions with
limited budgets. In view of this situation, the following section presents the ar-
chitecture for the creation of VDIs in a private open source cloud, OpenNebula.
The objective is to provide a low-cost virtual study environment in which labo-
ratory practices can be carried out, optimizing existing hardware and software
resources.
3 Proposed Architecture
When working on the development and design of VDIs as support tools for
teaching, it is essential to guarantee their correct functioning, since any error or
difficulty could cause a lack of motivation in the students and, therefore, learning
deficiencies that prevent the achievement of the objectives set in each subject. In
this way, the Fig 1 reflects the basic hardware and software components involved
in the deployment of VDIs, which can be summarized in 7 elements:
1. User access: access to each VDI is via any device with an Internet connection.
To do this, it is essential to have credentials (username and password), which
have been previously specified by the VDI administrator. This is due to
the fact that the desks are personalized since each student has their own
activities such as laboratory practices on specific topics.
2. VDI Manager: is in charge of managing the desktops, so it has sufficient per-
missions to deploy and configure virtual machines with operating systems
and pre-installed software. In addition, you must have the minimum capabil-
ities to perform license management tasks, package update, and repository
addition. The objective is to provide each student with a tool with the con-
figuration required for the development of the activities proposed by the
teacher.
3. End Devices: are devices through which you can access, view, and interact
with VDIs. It should be noted that their processing or storage characteristics
are not of major importance, since all the tasks in execution consume the
resources of the server that hosts the private cloud.
4. Physical server: is the host for the installation of the private cloud, which is
divided into two virtual servers that will do the functions of front and node.
On the one hand, the front is involved in the installation of the Sunstone
OpenNebula interface (a GUI that simplifies typical administration oper-
ations) to manage the cloud through a Web browser. Meanwhile, the node
provides underlying resources such as processing, storage, RAM and network
for the deployment of virtual machines, ensuring their proper functioning.
Fig. 1. Conceptual operation of VDIs on private cloud.
– CPU: Each virtual CPU core (VCPU) assigned to a VDI must exist as a
physical CPU core. For example, for a workload of 10 VDIs with 2 CPUs,
OpenNebula will need 20 physical CPUs, which could be distributed over 2
servers with 10 cores each. At this point, it is important to note that the
CPU indicates the physical CPUs assigned to the virtual machine; while the
VCPU establishes the virtual CPUs that will be made available to the guest
operating system.
– Memory: its planning is simple, since, by default, there is no excess memory
in OpenNebula. However, it is advised as a good practice to account for 10%
of the hypervisor overload (this is not an absolute upper limit, it depends on
the hypervisor). Therefore, to maintain a workload of 10 virtual machines
with 2GB of RAM, 20GB of physical memory will be required. Also, the
number of hosts is important, because they will incur a 10% overload due
to virtualization. For example, 10 hypervisors with 10GB of RAM will con-
tribute 9GB each (10% of 10GB = 1GB). The golden rule is to have at least
1GB per core.
– Network: must be carefully designed to ensure the reliability of the Open-
Nebula infrastructure. The recommendation is to equip with two front-end
cards (public and service) and four for each node (private, public, service
and storage). In some cases, fewer cards may be required depending on the
storage and network configuration.
– Storage: will depend on the size assigned to VDI. While generally working
with 100 GB, OpenNebula allows you to add more capacity if necessary.
In this way, they have specific tools to elaborate practical proposals on mathe-
matical modeling and configuration of e-mail and VoIP servers (Voice /IP). The
experiment demonstrated that, considering the dimensioning aspects of cloud
services in non-dedicated hardware, the work developed in each practice and the
access of the 15 students simultaneously during 20 working hours distributed in
4 working hours during 5 days, an expected behavior of hardware and software is
achieved, as can be observed in Fig. 2. Therefore, it can be stated that OpenNeb-
ula supports the workload of the VDIs, exposing a window of the remaining 10%
for peaks with overload and sustainable performance. This means that 100% of
the storage and memory allocated to the laboratories is used, maintaining a 90%
occupation of the real or underlying resources Fig. 3.
Fig. 2. Conceptual operation of VDIs on private cloud, work on a VDI using mathe-
matical software
On the other hand, in the second stage, the calculators of the biggest cloud
service providers in the market were used: Amazon Web Services, Microsoft
CPU hours Allocated CPU
15
10
100/100
5
Allocated Memory
0
19/05/18 19/05/20
30 Real CPU
20
10
90/100
0
19/05/18 19/05/20
Real Memory
Disk MB hour
6000000
4000000 90/100
0
19/05/18 19/05/20
100/100
Fig. 3. Conceptual operation of VDIs on private cloud, use of real resources, allocated
resources and overall system health
Azure and Google Cloud, to compare the annual cost of deploying the hardware
and software requirements of the experimental infrastructure. Table 2 shows
the relative cost increase of the alternatives paid with respect to OpenNebula
(local). It is interesting to note that the cost per year in a local architecture
is a product of the acquisition of servers and administrative variables such as
power, maintenance and trained personnel. While you might think that buying
equipment only generates administrative costs, you should keep in mind that a
scalable system supports the local addition of non-dedicated machines that lend
resources, increasing the annual investment.
Table 2. Allocation of resources to laboratories for virtual machines.
Services Machines number Host O.S. RAM Memory Hard Disk Relative Cost
Local 2 CentOS 32 GB 3 TB 1 USD
AWS 2 CentOS 32 GB 2 TB 3.59 USD
Azure 2 CentOS 32 GB 2 TB 2.80 USD
Google 2 CentOS 32 GB 2 TB 2.71 USD
5 Conclusions
Acknowledgements