Evaluation by Simulation of the Diffusion Methods in the Cloud:
Based Network Architecture for Digital Open Universities
Boukar Abatchia Nicolas, Mahamadou Issoufou Tiado, Nassirou Adamou Hassane
and Ibrahim Ganaou Noura
Department of Mathematics and Computer Science, Research Team on Network and Tele-communication,
University of Abdou Moumouni, BP 10662 Niamey, Niger
Nigurg44@gmail.com
Keywords: Distance Education, GSM, DOUNG, QoS, Cloud Computing.
Abstract: The interconnection between the Internet and the telecommunication networks brings to the advent of the new
generation of digital open universities (DOUNG). That recent model was improved through many additional
works including the extension of its architecture from the Local Area Network (LAN) to the Internet and to
the GSM (Global System for Mobile communications) environment. This hybrid architecture leads to several
connections with the goal to achieve a good level of Quality of Service (QoS). One solution belongs to the
using of clouds with the issue of choosing a diffusion method adapted to this new context. In this paper, a
comparative study of flow distribution methods is conduct through dissemination issue and simulations. We
extend that work with the cloud contribution assessment including scales evaluation. All results for vertical
and horizontal scaling and for the unicast and multicast methods are produced and discussed.
1 INTRODUCTION
The DOUNG architecture (
Tiado and all, 2013
) is
based on the Internet and the GSM from the wired and
wireless LAN. The service extension to the GSM
leads to several architectural solutions. Authors of
document (
Tawayé and all, 2021)
suggest
implementing an integrated server model (ISM) by
hosting mirror server at the GSM side with crossing
the Internet network. The idea here is to use clouds in
the global architecture to achieve a good level of QoS.
To evaluate the gains of this new model, it becomes
important to consider the method of flow
dissemination and the clouds distribution. The
contribution assessment is conduct through the
criteria of the global resources consumption in the
data center and at the run time. The goal is to evaluate
the system load for QoS performance, and the
processing time with the resource scaling which
indicates the optimization of their consumption
2 THE INTEGRATED SERVER
MODEL
The ISM describes the DOUNG architecture with a
mirror server hosted by the GSM Service Provider
(GSP) in addition to other services chain. Learners
with cellular devices can access this server locally,
which in turn maintains a sufficiently fluid
connection with that of the backbone to make
operational the synchronous course monitoring mode.
The problem of crossing the Internet network is
resolved through models for ensuring a minimum
quality of service using VPN (Virtual Private
Network) (Zhengchun and Tongcheng, 2021), (Mohd
F. and all, 2021), (Yunxiao S., and all, 2021), or
MPLS (Multi-Protocol Label Switching) (Faycal B.
and Najib K., 2019), (Anju B. and V.P.Singh, 2016)
or with IntServ (Integrated Services) and DiffServ
(Differentiated Services) (Abdullah Y. and Tolga G.,
2014). The figure 1 below summarizes the
architecture of this extended model.
Figure 1: DOUNG architecture extension with ad hoc
network.
276
Nicolas, B., Tiado, M., Hassane, N. and Noura, I.
Evaluation by Simulation of the Diffusion Methods in the Cloud: Based Network Architecture for Digital Open Universities.
DOI: 10.5220/0011236900003269
In Proceedings of the 11th International Conference on Data Science, Technology and Applications (DATA 2022), pages 276-280
ISBN: 978-989-758-583-8; ISSN: 2184-285X
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
3 NEW ARCHITECTURE WITH
INCLUSION OF CLOUDS
Clouds are integrated in the initial model of the
DOUNG to overcome the problems of nodes lacking
in terms of storage and processing capacity such as
cellular devices. In further work, the suggestion of the
ATRS (Advanced Text Reading System) is intended
to strengthen this contribution (
Tiado and all, 2015). In
document (Tiado and all, 2021), the solution of the
blackboard image transmission with the voice of the
teacher directed the additional works towards the
transcription of the voice for its conversion into text
format to facilitate its transport through the Internet
network. At the entrance of the GSM network, this
text stream is reconverted into audio by the ATRS.
The using of the cloud helps to offload the GSP
mirror server and to specialize it in the repatriation of
data streams from the DOUNG server. In this section,
we extend the clouds exploitation with two
complementary levels in addition to the GSM internal
cloud. A first extension consists of using a private
cloud in the backbone of the DOUNG and a second
extension allows to use a public cloud to facilitate
access to learners via the Internet network. The
expected gain is to obtain better QoS both with the
course warehouse storage and with the use of cloud
resources (CPU Central Process Unit, RAM
Random Access Memory, etc.) during live sessions of
classes.
3.1 The Improved Model
The figure 2 shows the LAN (Local Area Network)
and the WLAN (Wireless LAN) standing as the
backbone of the DOUNG with the interconnection of
the GSP mirror server. The last two environments
host private clouds. The need of QoS in the link
between the backbone and the public cloud, same as
the link toward the GSP, find solution within the three
methods including VPN, MPLS or IntServ and
DiffServ. The GSP network is part of the WAN
(Wide AN) with the possible inclusion of WiMAX
technology and other technologies depending on the
networks evolution.
Figure 2: Model of the general architecture of the DOUNG
network
3.2 Clouds Distribution
The use of clouds by the DOUNG can be realized
through three architectures: the Internet public cloud,
the DOUNG backbone private cloud and the GSP
private cloud.
Local Private Cloud Integration
In this model, all services are hosted in a private cloud
deployed in the DOUNG main site. The cloud is
however accessible from the Internet. Lessons are
produced and delivered directly to learners from this
cloud.
Public Cloud Integration
In this complementary model, all DOUNG services
are hosted in a public cloud. Lessons are produced by
the DOUNG and conveyed in this environment. The
learners most advantaged by this architecture access
lessons from the cloud.
Integration of the Private Cloud in the GSM
Environment
The previously model is extended with the addition
of a private cloud within GSM networks. This cloud
hosts temporary storage services and is used for
traffic filtering as well as all DOUNG services to
facilitate their access by mobile devices.
Final Hybrid Architecture
Figure 3: Architecture pattern with hybrid clouds.
4 COMPARATIVE STUDY OF
FLOW DISTRIBUTION
METHODS
Several metrics can be used to measure the
performance of our architecture according to the
cloud QoS metrics. These include performance,
safety (reliability), and configuration. For
performance metrics, the available settings are
response time, processing time, service throughput,
data transfer rate, and latency. Dependability
parameters relate to availability, elasticity, reliability,
Evaluation by Simulation of the Diffusion Methods in the Cloud: Based Network Architecture for Digital Open Universities
277
time independence, resilience, and scalability.
Finally, configuration metrics have parameters
including virtual systems and location.
For the performance gains evaluation of our
architecture, in the first place, we are particularly
interested by the bandwidth and the rate of losses. The
first parameter is assimilated to the throughput of the
service included in the performance metrics. The
second parameter assimilated to the reliability of the
service belongs to the dependability metrics. These
parameters will help to determine the distribution
method best suited to our context among the three that
can be used.
4.1 Dissemination Issue
The new DOUNG network architecture allows a
learner to follow a course synchronously by
consuming a cloud service locally, via the Internet or
in the GSM. To study and evaluate the QoS gains of
this technology, we first carry out a comparative
study of multimedia stream broadcasting techniques
to determine the best suited to real-time course
monitoring. For this, we use the NS2 and CloudSim
simulators to test our scenarios presented in the table
1. In a general way and in a computer network, there
are three modes or methods of diffusion usable for the
routing of data flow:
Broadcast: the source server stream is intended for all
connected clients. This type of broadcast is used for
applications such as IPTV (Internet Protocol
Television).
Multicast: the source server stream is intended for a
specific group of clients.
Unicast: the client connected to the server receives its
own stream.
4.2 Description of Scenarios
It is possible to classify learners into three categories.
The first category concerns learners directly
connected to local domains (LAN, WLAN and ad
hoc). The second category concerns GSM learners
and the third is that of learners connected from an
Internet access provider. According to the objectives
set by the DOUNG, these learners benefit from
flexible access to the synchronous service of the
course. How-ever, a first level of analysis allows to
consider that the broadcast mode does not comply
with these objectives. Indeed, it does not offer
flexibility to learners for having control over the
synchronous flow of multimedia data. Therefore, we
proceed to an evaluation of the two multicast and
unicast methods through four scenarios. We calculate
for each case the bandwidth consumed and the loss
rate. We use VBR (Variable Bit Rate) traffic for this
in the range of [448 to 648] Kbps. The following table
presents the topologies used for the simulations by
varying the nature of the topology and the diffusion
methods. Likewise, this table specifies the direction
of the traffic as well as the available bandwidth and
the transmission delay of each network link.
Table 1: Topologies used for simulations of diffusion
methods.
4.3 Simulation Results
The following table gives the results obtained for the
four scenarios. It is followed by the corresponding
graph.
Figure 4: Bandwidth and loss rate of diffusion methods.
The previous results produced under NS2
demonstrate that the single-group multicast mode
despite its performance proves to be inefficient for its
non-conformity to reality and its lack of flexibility.
The second multi-group multicast mode is closer to
the reality of DOUNG with learners scattered around
the world, but it is handicapped by a high error rate.
Unicast mode produces less loss rate but generates
high band-width consumption and cannot be
implemented exclusively at the risk of overloading
the DOUNG server and causing an explosion in
bandwidth consumption. There-fore, after the
synthesis of these results, we adopted the hybrid
broadcast model for its flexibility (unicast and
multicast) allowing some learners in the DOUNG
domain to join multicast groups while a second
DATA 2022 - 11th International Conference on Data Science, Technology and Applications
278
category is connected in unicast depending on the
load supported by the server. This model allows to
reduce the network load and consequently improves
the QoS of the synchronous mode of course
monitoring.
5 COMPARATIVE STUDY OF
FLOW DISTRIBUTION
METHODS
We are interested here by the global resources
consumption (GRC) (CPU, RAM, Number of VMs:
virtual machines) in the data center and at the Run
Time of the Tasks (RTTk) defined for each scenario.
The first parameter is used as an indicator of the
system load in the cloud while the second is
integrated into performance related QoS metrics,
including processing time. This second group of
parameters is evaluated with the resource scaling
(increase) which indicates the optimization of their
consumption during the execution of each scenario.
The simulation is related to every cloud of the three
architectures presented above (LAN, Internet, GSM).
5.1 Evaluation of Scales
The main characteristics of a good QoS in a cloud are
scalability and elasticity. In this part, we study the
scaling and load-balancing models allowing
horizontal autoscaling (creation of new VMs) and
vertical (increase CPU, RAM, Bandwidth) while
managing the allocation of learners to the appropriate
servers. A readjustment function for this affection is
provided to optimize the use of resources and
guarantee the performance and reliability of the
services. These models are evaluated by simulation
with CloudSim which is a tool (Java Library) for
simulating cloud computing scenarios. It provides
classes allowing to describe for a cloud, the data
centres (Datacentre), the virtual machines (Vm), the
applications (Cloudlet or Tasks), the users, the
computational resources, the strategies for managing
these resources and the load distribution strategies.
This set allows to evaluate the performance of clouds
with new models within calculation and scheduling
algorithms
5.2 Simulation Scenarios and
Parameters
The objective of this simulation is to evaluate the
computing resources consumption (CPU, RAM,
Storage) and performance (execution time, response
time) for a better QoS of monitoring the DOUNG
course in synchronous mode. We propose different
scenarios highlighting vertical and horizontal scaling
strategies to deter-mine resource utilization and task
execution time through the following combinations:
Scenario 1: does not apply scaling;
Scenario 2: vertical scaling of the CPU;
Scenario 3: RAM vertical scaling;
Scenario 4: horizontal scaling of VM;
Scenario 5; CPU and RAM vertical scaling;
Scenario 6: VMs horizontal scaling and the
CPU vertical scaling;
Scenario 7: horizontal scaling of VMs and
vertical scaling of RAM;
Scenario 8: the VMs horizontal scaling with
the CPU+RAM vertical scaling;
The global parameters of the simulation are the
following:
15 physical machines of 2 GB of RAM, 100 GB of
hard disk, 4 CPUs and 10 GB of bandwidth with a
processor of 1000 MIPS (Millions of Instructions Per
Second);
At start-up, 4 virtual machines are created of 1 GB of
RAM, 10 GB of hard disk, 2 CPU and 10 GB of
bandwidth with a processor of 1000 MIPS each. They
are allocated to 100 tasks of size 2000 MI (Million
Instructions) each.
The scenario evolves with the addition of 20 tasks of
size 2000 MI each in the interval [0, 50] seconds
during the simulation.
The time-sharing strategy is used for VMs and tasks.
Horizontal scaling creates 12 new virtual machines.
The results illustrated in the figure 4 indicate the
overall use of resources (CPU, RAM, Number of
VMs) in the data center and the task execution time.
Figure 5: Summary of scaling results.
The previous figure allows to assess the performance
gains in the eight (8) scenarios. With the increase in
the system load previously described, the simulation
Evaluation by Simulation of the Diffusion Methods in the Cloud: Based Network Architecture for Digital Open Universities
279
results show that CPU vertical scaling is very
efficient for good performance and good QoS in the
cloud with run time equal to 50 seconds. This is
complemented by horizontal scaling in the VM-CPU
and VM-CPU-RAM scenarios in which the execution
time increases to 46 seconds. However, it also
requires the availability of a minimum of resources
for its operation. It should also be noted that the
vertical scaling of the RAM does not act enough on
the execution time which, as we can notice, re-mains
intact in the scenarios which involve it as in the same
scenario where it is not used (example result without
scaling = result with RAM scaling = 60 seconds).
6 CONCLUSION
The evolution of the DOUNG model leads to consider
the constraints of facilitating access and ensuring an
appreciable level of QoS. To solve the QoS problem
of crossing the Internet for the link between the LAN
and the GSM, the ISM model has been proposed
using VPN, MPLS or the association of the IntServ
and DiffServ protocols. One of the contributions
developed in this paper relates to the inclusion of
cloud technology. Thus, to evaluate the performance
gains in QoS that can be achieved, we have carried
out a comparative study of the methods of streaming
flows and defined simulation scenarios. It appears
from this study that the hybrid broadcasting method
(unicast and multicast) offers flexibility of choice to
learners, reduces the network load, and consequently
improves the QoS of the synchronous mode of course
monitoring. As a prelude to the prospect of evaluating
by simulation the contribution of the clouds, an
evaluation of the scaling was carried out with new
scenarios and simulation parameters. A next step will
consist in comparing the performance gains obtained
between the architectures including and excluding the
use of clouds.
REFERENCES
M. I. Tiado, H. Saliah-Hassane (2013). Cloud-Computing
based architecture for the advent of a New Generation
of Digital Open Universities in m-learning. ICEER13.
B. A. Tawayé, M. I. Tiado, S. Abdoulwahabou, M.
Harouna, I. G. Noura (2021). Models of Quality of
Service (QoS) in the GSM environment of the New
Generation of Digital Open Universities (DOUNG).
International Journal of Wireless Networks and Com-
munications.
Zhengchun Z., Tongcheng H. (2021). Open VPN
Application in COVID-19 Pan-demic. International
Conference on Advances in Optics and Computational
Sciences, Journal of Physics.
Mohd F., Mohamad A, Naginder K., Iman H. (2021).
SafeSearch: Obfuscated VPN Server using Raspberry
Pi for Secure Network. Journal of Computing Research
and Innovation (JCRINN).
Yunxiao S., Bailing W., Chao W., Yuliang W. (2021). On
Man-in-the-Middle Attack Risks of the VPN Gate Relay
System. Hindawi Security and Communication
Networks.
Faycal B., Najib K. (2019). Novel Software-Defined
Network Approach of Flexible Network Adaptive for
VPN MPLS Traffic Engineering. International Journal
of Advanced Computer Science and Applications.
Anju B., V.P.Singh (2016). Proposal and Implementation
of MPLS Fuzzy Traffic Monitor. International Journal
of Advanced Computer Science and Applications.
Abdullah Y., Tolga G. (2014). Cloud Technology and
Performance Improvement with IntServ OVER DiffServ
for Cloud Computing. International Conference on
Future Internet of Things and Cloud, IEEE Computer
Society.
M. I. Tiado, A. Idrissa, D. Karimou (2015). Improved Text
Reading System for Digital Open Universities.
International Journal of Advanced Research in
Artificial Intelligence.
M. I. Tiado, I. G. Noura, C. I. Hussein, H. G. Souleymane,
H. M. Mahamadou (2021); Quality of Service
Evaluation with DSR (Dynamic Source Routing)
protocol in the classroom ad hoc network of the New
Generation of Digital Open Universities (DOUNG).
16th International Conference on Information
Assurance and Security, Springer Nature Switzerland.
DATA 2022 - 11th International Conference on Data Science, Technology and Applications
280