Docsity
Docsity

Prepara tus exámenes
Prepara tus exámenes

Prepara tus exámenes y mejora tus resultados gracias a la gran cantidad de recursos disponibles en Docsity


Consigue puntos base para descargar
Consigue puntos base para descargar

Gana puntos ayudando a otros estudiantes o consíguelos activando un Plan Premium


Orientación Universidad
Orientación Universidad


CONTROLADOR ONOS PARA UNA RED SDN, Guías, Proyectos, Investigaciones de Redes de Computadoras

Comandos para la configuracion del controlador ONOS, en SDN, redes definidas por software

Tipo: Guías, Proyectos, Investigaciones

2019/2020

Subido el 05/12/2020

nicolas-arotoma-pravia
nicolas-arotoma-pravia 🇵🇪

1 documento

1 / 12

Toggle sidebar

Esta página no es visible en la vista previa

¡No te pierdas las partes importantes!

bg1
International Journal of Computer Science & Information Technology (IJCSIT) Vol 10, No 5, October 2018
DOI: 10.5121/ijcsit.2018.10503 21
S
IMULATION
O
F
S
OFTWARE
D
EFINED
N
ETWORKS
W
ITH
O
PEN
N
ETWORK
O
PERATING
S
YSTEM
A
ND
M
ININET
Antonio Cortes
Department of Computer Engineering, University of Panama, Panama City, Panama
A
BSTRACT
With the emergence of recent te chnologies in the field of computer network, traditional infrastructure in t he
field of networks have become obsolete and incompatible with respect to the new architectures of open
networks that emerge with force. This is how software-defined netw orks emerge by enabling cloud
computing ecosystem, enterprise data centers, and telecommunications service providers. The major
contribution of this paper is the simulation of an ecosystem based on a software defined network by making
use of certain types of networks topologies and using the virtualization of the open network operating
system (ONOS) and Mininet as a network em ulator.
K
EYWORDS
Computer Network, Software-defined Networks, Network Topologies, Simulation, Open Network Operating
System.
1.
I
NTRODUCTION
The infrastructure of the traditional networks was developed and implemented in such a way the
flow control and the routing oversaw the devices of the network, thus allowing static structure [3],
hierarchical and dependent on the network architecture. This makes the network a complex
scenario, because there is to accommodate the structure of the network to the needs of users,
taking into consideration the policies of a network [1] and the growth of these, since they are
problems those who face these designs.
Currently, with the emergence of social networks, smart devices and cloud computing, the
various network topologies that make up these infrastructures tend to become saturated and
consume a large bandwidth due to the electronic components of these innovative technologies. In
turn, the benefits of cloud computing and virtual storage are being limited by networks since the
implementation of new networks prevents them from meeting the new needs demanded by them
and demands from the operators’ market. In turn, the above gives rise to the emergence of a new
network model that improves the capabilities in these networks called Software Defined
Networks (SDN) [2].
In this way, Software Defined Networks are a network design architecture in which the control
layer and the data layer within a network are separated. It separates the hardware management
software from the network and transfers the control to other devices called controllers that
convert the data traffic and control the network into a centralized service [4]. At the same time,
the emergence of these networks is to cover a need related to the deficiencies of current networks
Electronic copy available at: https://ssrn.com/abstract=3286869
pf3
pf4
pf5
pf8
pf9
pfa

Vista previa parcial del texto

¡Descarga CONTROLADOR ONOS PARA UNA RED SDN y más Guías, Proyectos, Investigaciones en PDF de Redes de Computadoras solo en Docsity!

DOI: 10.5121/ijcsit.2018.10503 21

SIMULATION OF SOFTWARE DEFINED NETWORKS

WITH OPEN NETWORK OPERATING SYSTEM AND

MININET

Antonio Cortes

Department of Computer Engineering, University of Panama, Panama City, Panama

ABSTRACT

With the emergence of recent technologies in the field of computer network, traditional infrastructure in the field of networks have become obsolete and incompatible with respect to the new architectures of open networks that emerge with force. This is how software-defined networks emerge by enabling cloud computing ecosystem, enterprise data centers, and telecommunications service providers. The major contribution of this paper is the simulation of an ecosystem based on a software defined network by making use of certain types of networks topologies and using the virtualization of the open network operating system (ONOS) and Mininet as a network emulator.

KEYWORDS

Computer Network, Software-defined Networks, Network Topologies, Simulation, Open Network Operating System.

1. INTRODUCTION

The infrastructure of the traditional networks was developed and implemented in such a way the

flow control and the routing oversaw the devices of the network, thus allowing static structure [3],

hierarchical and dependent on the network architecture. This makes the network a complex

scenario, because there is to accommodate the structure of the network to the needs of users,

taking into consideration the policies of a network [1] and the growth of these, since they are

problems those who face these designs.

Currently, with the emergence of social networks, smart devices and cloud computing, the

various network topologies that make up these infrastructures tend to become saturated and

consume a large bandwidth due to the electronic components of these innovative technologies. In

turn, the benefits of cloud computing and virtual storage are being limited by networks since the

implementation of new networks prevents them from meeting the new needs demanded by them

and demands from the operators’ market. In turn, the above gives rise to the emergence of a new

network model that improves the capabilities in these networks called Software Defined

Networks (SDN) [2].

In this way, Software Defined Networks are a network design architecture in which the control

layer and the data layer within a network are separated. It separates the hardware management

software from the network and transfers the control to other devices called controllers that

convert the data traffic and control the network into a centralized service [4]. At the same time,

the emergence of these networks is to cover a need related to the deficiencies of current networks

because, by reducing the hardware necessary for assembly, facilitates the reuse of hardware by

facilitating the management of network control elements and makes it simpler in its configuration

process and reduces management time for administrators, speeding up the deployment of

applications, services and infrastructures.

Therefore, SND networks allow to ensure that network engineers and administrators respond

quickly to changes in business by centralizing the control console, by improving the services on

the network by making them more dynamic, economical and scalable avoiding management at a

low level. Similarly, SDN networks are composed of three essential parts. On the one hand, we

have the controller that oversees managing the network and telling the rest of the devices how to

handle the traffic on the network. The Southbound APIs is a software like the OpenFlow protocol,

responsible for managing the communication between the controller and the devices. While the

Northbound APIs is responsible for establishing communication with applications and business

logic. In this way, for a network to be functional it is required that the devices of this have

incorporated a firmware with OpenFlow or another similar protocol. This means that SDN

network flexibly manage the devices through the controller, which is responsible for recognizing

the topology of the network, thus enabling better management of traffic loads in a flexible and

efficient manner, giving priority to processes, applications and services.

In this paper, we have considered the use of the Open Network Operating System (ONOS) and

the Mininet network emulator to be able to simulate the Software Defined Network (SDN)

ecosystem and their respective network topologies. It makes use of a graphical interface for the

virtual machine VirtualBox of Oracle, version 5.2.8 r121009 (Qt 5.6.2) in which ONOS is

installed with its respective Mininet emulator, the OpenFlow controller and the Ubuntu Operating

System 16.0.4.3.

The organization of this paper is as follows. In Section 2. Methods and Materials refers to the

inputs used in the preparation of this paper. In turn, Section 3. Results and Discussion, the

analysis of SDN is illustrated through ONOS and Mininet simulation tool. Section 4. Explains the

final considerations and details the references used in this paper.

2. METHODS AND MATERIALS

A systemic process is carried out that allows the ONOS, Mininet and revision of the

documentations from the various positions proposed by each one of the different authors.

2.1. OPEN NETWORK OPERATING SYSTEM (ONOS)

The Open Network Operating System is an open source distributed SDN control platform,

developed by the Open Networking Lab (ON, Lab) [5], and sponsored by some of the leading

companies and academic institutions. In comparison with Open-DayLight [6], the ONOS Project

is specifically oriented to ISP (Internet Service Provider) networks, facilitating high availability

and scalability, due to its distributed architecture. The identification of the network topologies, as

well as our results obtained in the SDN simulations, are carried out using the version of ONOS

1.13.1 and the Mininet network emulator.

On the other hand, to establish the design of the SDN network, as well as the number of

controllers or data switches, number of switches, identification of physical links through which

the exchanged packets travel between the control plane at the application level, not only the

ONOS platform was used, but two complementary approaches were also adapted. On the one

hand, we use the SDN Mininet network emulator [7], to create a test network topology, which is

shown in Figure 1., and we proceed to execute an instance of ONOS with fwd (Simple reactive

2.3. MININET

An emulator is software that allows programs to be run on a different platform than the one

originally designed. Unlike a simulator, it only reproduces the behavior of the program while an

emulator accurately models a device that can be compared with the original hardware.

MiniNet [9] is one of the first emulators explicitly developed to support SDN, by allowing the

efficient execution of small-scale networks with artificial traffic on computers that are not

necessarily powerful, its license is free and permissive (BSD - Berkeley Software Distribution).

In addition, implementing the network with a large number of network devices is very difficult

and expensive. Therefore, to overcome these problems, the virtual mode strategy has been carried

out in order to create prototypes and emulation of technological networks using the MiniNet

emulator. Its operation is carried out through a single Linux kernel and uses virtualization in order

to emulate a complete network using only a single system. However, the node created, as well as

the switches, routers and links are real elements although they are created by software [10].

The goal of MiniNet is to create virtual networks, running nodes, network cores and virtualized

network devices simply and quickly through a simple feature host, with an open and free

environment such as Linux. In turn, it has the ability to emulate different types of elements of the

network, such as: nodes, layer 2 switches, layer 3 routers and links.

Some features that led to the creation of MiniNet are:

  • Flexibility, that is, new topologies and new features can be configured through the use of

software, by implementing common programming languages and operating systems.

  • Applicability, allows correct applications made in prototypes. They should also be able to be

used in real networks based on hardware without any change in the source codes.

  • Interactivity, responsible for managing and executing the simulation of the network, so it

must occur in real time as if it were happening with a real network.

  • Scalability, prototyping must be scaled in large networks with hundreds or thousands of

switches on a single computer.

  • Realistically, the behavior of the prototype must represent the real behavior of time with a

high degree of confidence, so the application and protocol stacks should be usable without

any code modification.

  • Shareable, prototypes created should be easily shared with other collaborators, who can then

execute and modify the elements [11].

2.3.1 MININET WORKFLOW

Mininet has the capabilities that allow researchers or network programmers to create a new

software-defined network in a prototype and simple way, with the ability to interact, customize

and share, and provide a way to be executed on the hardware.

 Creating a network.

You can create a network in MININET with a single command;

$ sudo mm --switch = ovsk, protocols = OpenFlow13 --controller = remote --topo = tree, depth =

4, fanout = 2 --ipbase = 172.17.0.2 / 20

Create a virtual network of four switches or switches connected in a tree topology to two hubs

and 20 nodes, each node configured for the corresponding switch, so there would be a distribution

of 5 nodes per switch, as show in Figure 2.

Figure 2. Network topology in Tree and its components

 Interacting with a network

In Mininet, the entire virtual network can be controlled, and managed from a single console, for

example, the CLI command.

Mininet> h1 ping -c3 h

It is used to send a ping to node h2 from node h1.

Mininet> nodes

View the list of available nodes.

Mininet> help

Allows you to see a list of available commands.

Dpctl: controls and edits flow tables.

Iperf: Test the TCP speed.

 Customize a network.

Custom networks with a few lines of Python can be created with the Mininet API. For example, ~

/ mininet / custom / topo-4sw-20host.py

These few lines of Python create a virtual network of twenty nodes connected through virtual

links to four switches.

 Share a network.

Mininet allows you to share the created, an image of VM to other researchers with the

purpose of running, evaluating or modifying it.

3. RESULTS AND DISCUSSION

Initially, the Open Network Operating System (ONOS) was activated, through the icon

named Setup ONOS Cluster. Subsequently, the network topology configuration called

Spine Leaf Topology or tree network topology is activated in such a way that activates all

the components of this topology. At the same time, the virtualization process of the

duplex, represented by the equation A↔ B, which means that the data bits, zeros and

ones, travel simultaneously between two or more devices.

In Table 3. , we identify the total of nodes or host that are 20, with the following

parameters.

Table 3. Host.

Host Host Id Mac Address Location 10.0.0.1 00:00:00:00:00:01 / None 00:00:00:00:00:01 of:000000000000000b/ 10.0.0.2 00:00:00:00:00:02 / None 00:00:00:00:00:02 of:000000000000000b/ 10.0.0.3 00:00:00:00:00:03/ None 00:00:00:00:00:03 of:000000000000000b/ 10.0.0.4 00:00:00:00:00:04 / None 00:00:00:00:00:04 of:000000000000000b/ 10.0.0.5 00:00:00:00:00:05 / None 00:00:00:00:00:05 of:000000000000000b/ 10.0.0.6 00:00:00:00:00:06 / None 00:00:00:00:00:06 of:000000000000000c/ 10.0.0.7 00 :00:00:00:00:07 / None 00 :00:00:00:00:07 of:000000000000000c/ 10.0.0.8 00:00:00:00:00:08 / None 00:00:00:00:00:08 of:000000000000000c/ 10.0.0.9 00:00:00:00:00:09 / None 00:00:00:00:00:09 of:000000000000000c/ 10.0.0.10 00 :00:00:00:00:0A / None 00 :00:00:00:00:0A of:000000000000000c/ 10.0.0.11 00:00:00:00:00:0B / None 00:00:00:00:00:0B of:000000000000000d/ 10.0.0.12 00:00:00:00:00:0C / None 00:00:00:00:00:0C of:000000000000000d/ 10.0.0.13 00:00:00:00:00:0D / None 00:00:00:00:00:0D of:000000000000000d/ 10.0.0.14 00 :00:00:00:00:0E / None 00 :00:00:00:00:0E of:000000000000000d/ 10.0.0.15 00:00:00:00:00:0F / None 00:00:00:00:00:0F of:000000000000000d/ 10.0.0.16 00:00:00:00:00:10 / None 00:00:00:00:00:10 of:000000000000000e/ 10.0.0.17 00 :00:00:00:00:11 / None 00 :00:00:00:00:11 of:000000000000000e/ 10.0.0.18 00:00:00:00:00:12 / None 00:00:00:00:00:12 of:000000000000000e/ 10.0.0.19 00:00:00:00:00:13 / None 00:00:00:00:00:13 of:000000000000000e/ 10.0.0.20 00:00:00:00:00:14 / None 00:00:00:00:00:14 of:000000000000000e/

In Table 3. , we observe the IP addresses assigned to each node, as well as the respective

identifier and mac address associated with each node. In the range of IP addresses

10.0.0.10 to 10.0.0.15, hexadecimal values are assigned to both the identifier and the mac

address of the respective nodes. In the location, each IP address of each node has the

respective switch assigned by a letter and the node number.

On the other hand, in table 4. , we obtain the processing of packets at different times of

activation of the topology of the network.

Table 4. Packets Processors.

Priority Packets Average (MS)

Priority 0

P1 P2 P3 P4 P5 A1 A2 A3 A4 A

Priority 1

Priority 1

In Table 4. , we can see that packet processing will have three priority types, 0, 1 and 1,

respectively, established for this tree network topology. We did tests in the network with different

time intervals of connection and disconnection of the same, to observe the behavior at the level of

the processing of packages and the average of them. In addition, it was possible to observe the

bits per second at the ports level, the packets per second at the level of the statistics of the ports

and bytes at the level of the data flow statistics.

In Figure 3a.,and Figure 3b. , we can observe the relationship between the different

priorities with respect to the packages and the averages, respectively.

4. CONCLUSIONS

Figure 3a. Priority vs Packets Figure 3b. Priority vs. Average

In Figure 3a. , we can see that at the level of priority 1, in gray, the maximum number of

packets sent in the network reaches its maximum trajectory in 1576 packets, while the

lowest is in 36 packets. Between P3 and P4, the oscillation tends to increase with a

difference of 136 packets ((P4 = 676) - (P3 = 540) = 136). In the case of Figure 3b. , it is

observed that the maximum average is reached when A2 = 30.17557 milliseconds (ms) in

priority 0, in blue. The lowest average is given when A2 = 0.00300 milliseconds (ms) in

priority 1, in gray. The latter is due to a decrease in the bandwidth of the network or

otherwise a disconnection in some of its links due to a fault has occurred.

In the same way, other performance metrics were taken into account to evaluate the

performance of the network. These parameters are related to the bandwidth which is

analyzed through each of the ports located in the controllers and switches, respectively, as

we can see in Table 5.

Table 5. Ports for controllers and switches.

Device Enabled ID Speed Type

Spine-

Spine-

False Local 0 Copper True 1 10000 Copper True 2 10000 Copper True 3 10000 Copper True 4 10000 Copper

Leaf-

Leaf-

Leaf-

Leaf-

False Local 0 Copper True 1 10000 Copper True 2 10000 Copper True 3 10000 Copper True 4 10000 Copper True 5 10000 Copper True 6 10000 Copper True 7 10000 Copper

However, and with all the above, the network topology proposed in this work is

compared with the tree topology with the MPLS topology (Multiprotocol label

switching), since it is a very popular method used for traffic control and the creation of

virtual private networks (VPNs). This method known as "tunnel-less", is a bit

complicated to understand because it lacks a point-to-point connection. In Figure 4, the

MPLS topology is shown.

Figure 4. MPLS Topology

This scenario, which is presented in Figure 4, allows us to observe that at the level of

communications infrastructure, what dominates the most are switches interconnected

with fiber optic links or, ultimately, copper. It is important to mention that in this type of

topology there is a drop in logical links due to the creation of virtual private networks.

The existence of subnets composed of nodes is not unlike the tree topology that does

exist. In this type of scenario, clients connect to the backbone of the network through

multiservice links, which provides high transport speed and connectivity.

In addition, it is important to highlight the data packet flow that occurs in this type of

topology with respect to the tree topology, as shown in Figure 5a-b.

Figure 5a. Flow packets in Tree Topology Figure 5b. Flow packets in MPLS Topology

It is observed, in Figure 5a. and 5b. , respectively, in green, the flow of data packets is

more intense in the MPLS topology than in the Tree topology. Other reasons, which may

lead to the fall of the logical links in the MPLS topology is the means by which the data

is routed since it is copper.

4. CONCLUSIONS

In this article it is proved that the Open Network Operating System (ONOS) and the

Mininet network emulator allow researchers in the field of networks to virtualize a

network topology in tree and any other, with all its main actors as are: the switches, the

controllers, the nodes connected to each switch, establish the IP address for the network

in general and the IP addresses for each subnet, among other relevant aspects. Similarly,

it could be observed in the tree network topology, when the links between different nodes

break or fall or between the controller and a node. This situation is represented in the

virtual network, as discontinuous lines in red. It was observed how the bits per second are

transmitted between the various ports of each device. This process of data transmission

was reflected in the network in green.

On the other hand, a comparison process is carried out at the level of new network

performance metrics between which the bandwidth and throughput are considered. Both

metrics are analyzed taking into account the ports enabled in each switch and controller,

packets received and transmitted at the level of these devices. Also, a comparison is made

at the level of the proposed network topology against the MPLS topology, among which

the most outstanding is the flow of data packets between both topologies and that the

MPLS topology does not have sub-networks formed by nodes but interconnections

between different linked switches by means of a communication channel formed by

copper or optical fiber.

It is important to point out that where the experiments were carried out, the Hardware and

the Software used were installed in a desktop-like technological infrastructure. We used a

Lenovo Laptop with a 64-bit Operating System (Windows 10), an Intel Celeron 3205 U

processor with a speed of 1.50 GHz and 4.00 GB RAM. In the same way, it was installed

and made use of a virtual machine, Oracle VirtualBox, where ONOS and Mininet were

installed, thus optimizing the infrastructure and improving the performance of those

elements that were available at that time.

As future lines of research, ONOS would be used with other types of network emulators

to analyze the behavior of different topologies of more extensive networks and to start

with a comparison process, thus allowing us to observe and compare new results in this

type of ecosystem.

REFERENCES

[1] Carvalho, L. Fernando, Abrão. T, Mendes L. de Souza, ProençaJr. L. Mario, An Ecosystem for

Anomaly Detection and Mitigation in Software-defined Networking, Expert Systems with Applications (2018), doi: 10.1016/j.eswa.2018.03.

[2] F. Keti and S. Askar, "Emulation of Software Defined Networks Using Mininet in Different Simulation Environments," 2015 6th International Conference on Intelligent Systems, Modelling and Simulation, Kuala Lumpur, 2015, pp. 205-210, doi: 10.1109/ISMS.2015.