DEVS-M: A discrete event simulation framework for MANETs

DEVS-M: A discrete event simulation framework for MANETs

Accepted Manuscript Title: DEVS-M:A discrete event simulation framework for MANETs Author: Fatih C ¸ elik PII: DOI: Reference: S1877-7503(15)30044-2 ...

3MB Sizes 1 Downloads 107 Views

Accepted Manuscript Title: DEVS-M:A discrete event simulation framework for MANETs Author: Fatih C ¸ elik PII: DOI: Reference:

S1877-7503(15)30044-2 http://dx.doi.org/doi:10.1016/j.jocs.2015.11.012 JOCS 439

To appear in: Received date: Revised date: Accepted date:

21-7-2015 29-10-2015 27-11-2015

Please cite this article as: Fatih C ¸ elik, DEVS-M:A discrete event simulation framework for MANETs, Journal of Computational Science (2015), http://dx.doi.org/10.1016/j.jocs.2015.11.012 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Biographies (Text)

Author Biography Dr. Fatih Çelik

Ac

ce pt

ed

M

an

us

cr

ip t

Fatih Çelik is Assist. Professor at Sakarya University, Turkey. His research topics include DEVS theory, multi-formalism modeling, parallel and distributed simulation, modeling and simulation of large-scale networks, distributed systems management, biologically-inspired optimization schemes and Wireless sensor network. His main research interest lies in parallel and distributed simulation and routing protocols for wireless sensor network.

Page 1 of 27

Ac ce p

te

d

M

an

us

cr

ip t

*Biographies (Photograph)

Page 2 of 27

*Highlights (for review)

The Highlights

ip t

cr us an M ed



ce pt

 

Developed simulation framework was developed by adapting DEVS-Suite simulator tool using DEVS formalism DEVS-M simulator provides close findings to expected results in the real life. DEVS-M simulator was found to produce lower results than ns-2 simulator in terms of “memory consumption for per connection” but higher than other simulators. The simulator kept its stability in terms of “memory consumption for all connections” and the increase in the number of nodes did not affect memory consumption.

Ac



Page 3 of 27

ip t

DEVS-M:A discrete event simulation framework for MANETs Fatih C ¸ elik∗

us

cr

Department of Computer Engineering, Sakarya University, Esentepe Campus/ Serdivan/Sakarya, Turkey

Abstract

an

Modelling and simulation is important for Mobile Ad-hoc NETworks (MANETs). Studies in literature show that simulation tools developed for cable networks have been transformed into MANET tools instead of developing MANETs spe-

M

cific simulation tools. In this paper, simulation framework for MANETs is designed by using the DEVS-Suite simulator tool based on DEVS formalism and used for the simulation of cable networks. Besides, specific wireless node

d

architecture for the nodes that forms MANET, a topography model to check

te

the velocity and movements of the nodes, packet models roaming in MANET and visualization area to observed simulation works is developed for simula-

Ac ce p

tion framework. A coupled model is generated by combining these two models. Also, An ant colony based load balancing scheme is developed to test the model. As a result, the simulation tool developed for MANET aims to keep memory consumption steady even when traffic and node intensity increases in simulators. The study also aims to undertake load balancing tests to obtain results compatible with the values in literature. Keywords: Distributed Network Simulator, MANET, DEVS, DEVS-M

∗ Corresponding

author Email address: [email protected] (Fatih C ¸ elik )

Preprint submitted to Journal of LATEX Templates

May 27, 2015

Page 4 of 27

1. Introduction

ip t

At first, this level of development for MANETs was not foreseen. The developments necessitated modeling and simulation tools to verify and validate the protocols developed for MANETs. The modelling and simulation tools developed for this purpose allows testing of protocols designed for MANETs.

cr

5

Tools used in the literature for MANETs modeling and simulation were de-

us

veloped as MANETs adaptation of tools generated to model cable computer networks [1] . Therefore, while the links in these tools were adapted to antenna models,cable nodes were adapted to wireless nodes. One of the difficulties observed in wireless nodes is the placement of the nodes since wireless nodes are

an

10

placed at networks randomly in contrast to cabled nodes. In order to overcome

M

this problem, topology generators are adapted to fulfill wireless node placement requirements. Another difficulty faced in wireless nodes is the mobility of network nodes whereas no mobility is possible in cable nodes. Hence, the simulator that will be developed requires the addition of a tool that does not exist in cable

d

15

node. The developed tool will be able to check both location and velocity infor-

te

mation for wireless nodes. Instantaneous change in the location of the mobile node during mobility should be identified by the simulator depending on its

Ac ce p

velocity.

20

There are various simulators for MANETs in the literature and the most

commonly used simulator tool is ns-2 [2]. Some errors occur in the transmission range during node placement caused by PRNG (pseudo random number generator) and this situation shows the importance of topology generator. Another commonly used simulator is OPNET [3]. OPNET is a simulator with commer-

25

cial limitations and therefore generates a high accuracy expectation [1] which does not scientifically allow the formation of an experimental framework. There are also different simulator tools written by users themselves regarding specific experiments. In this paper, we developed a MANET simulator which based on DEVS [4].

30

This simulator has the ability to work on distributed and parallel architectures

2

Page 5 of 27

using discrete event based DEVS approach [5]. This MANET simulator called DEVS-M is developed using DEVS-Suite [6].

ip t

The MANET simulation tool can serve parallel and distributed architectures.

As opposed to other simulators, creation of scenarios and monitoring the results of these scenarios are easier with the help of these developed tools. It is also

cr

35

possible to use it as a web-based MANET simulator tool in distance education

us

since it is independent of platforms and it is an open source. It is different from other simulators since it does not change the memory consumption even when the load in simulators increases. Ant colony based [7] AntODV [8] load balancing scheme based on AODV [9] approach was developed to present the superior characteristics of the simulator.

an

40

Section 2 briefly reviews DEVS formalism. Section 3 explains the atomic

M

node model, topography atomic model, packets, coupled model, topology generator developed for MANET simulator, modelling processes and architecture 45

of visualization area tools. Section 4 provides information on the simulation pa-

d

rameters necessary for the verification and validation for the developed MANET

te

simulator tool. Section 5 graphically presents the results obtained in the experiments and provides performance assessments for them. Section 6 provides

Ac ce p

conclusions.

50

2. Discrete event modelling (DEVS) The modeling and simulation method used in the current study was first

presented for the modeling and analysis of discrete event systems by Dr. Bernard P. Zeigler in 1976 in his book titled ”Theory of Modeling and Simulation”. Discrete Event System Specification (DEVS) defines system behavior in two

55

different levels: atomic and coupled DEVS. At the lowest level, DEVS defines the autonomous behavior of a discreet event system such as the transitions between sequential cases, how the system responds to an external input (events) and how it evaluates the output (events) [10]. An atomic DEVS model is defined as following.

3

Page 6 of 27

60

M =< X, Y, S, ta, δint , δext , λ > X is the set of input events. Y is the set of output events. S is the set of

ip t

sequential states. ta : S → T ∞ is the time advance function which is used to determine the lifespan of a state. δext : Q × X → S is the external transition

65

cr

function which defines how an input event changes a state of the system, where

Q = {(s, te )|s ∈ S, te ∈ (T ∩ [0, ta(s)])} is the set of total states, and te is the

us

elapsed time since the last event. δint : S → S is the internal transition function which defines how a state of the system changes internally (when the elapsed time reaches to the lifetime of the state). λ : S → Y φ is the output function

70

an

where Y φ = Y ∪ {φ} and φ 6∈ Y is a silent event or an unobserved event. This function defines how a state of the system generates an output event (when the elapsed time reaches to the lifetime of the state)[4].

M

Coupled DEVS defines a system as a network of components at a higher level. The components may be atomic DEVS models or coupled DEVS models. Couplings show how components affect each other. Output events of a component may be the input events of another component. It is possible to design an

d

75

te

atomic DEVS for each coupled DEVS and atomic or coupled DEVS model can be displayed by an atomic DEVS [11]. Hierarchical modelling structure is sup-

Ac ce p

ported since coupled DEVS can have other coupled DEVS components [12].A coupled DEVS model is defined as following.

80

N =< X, Y, D, {Mi }, Cxx , Cyx , Cyy , Select > X is the set of input events. Y is the set of output events. D is the name set of

85

sub-components {Mi } is the set of sub-components where for each i ∈ D, Mi can S be either an atomic DEVS model or a coupled DEVS model. Cxx ⊆ X × i∈D Xi S S is the set of external input couplings. Cyx ⊆ i∈D Yi × i∈D Xi is the set of S internal couplings. Cyy : i∈D Yi → Y φ is the external output coupling function Select : 2D → D is the tie-breaking function which defines how to select the event from the set of simultaneous events. Advantages of DEVS approach are strong connections between components, hierarchical design, event-based simulation, object oriented adaptation, reduced

90

design time, more developed tests, higher quality models, easier experimenting 4

Page 7 of 27

opportunities, autonomous opportunities for working in parallel/real time, ease of verification and validation, interoperability and reuse, modelling by using

ip t

more than one method and high performance [13]. This study presents the performance achievements of DEVS method especially on parallel and distributed systems such as networks by utilizing the advantages cited above.

cr

95

us

3. The components of the DEVS-M

Each node is designed in a way to process the packets in the formed topology and to route these packets to desired directions. Nodes are atomic models that

100

an

can connect to and communicate with the nodes included in the coverage area in the system as in reality. Behavioral characteristics of the nodes include the band width to process the traffic, processing velocity and limited buffering size

M

to process traffic. Network components of various capacities can be formed by changing the defined characteristics and different network scenarios can be

105

d

developed [12]. The wireless network model has two different types of atomic model, a network coupled model and experimental frame. The node atomic

te

model and topography atomic model form coupled node models. Furthermore, experimental frame model has also event generator and event transducer atomic

Ac ce p

models (see Figure 1). Every wireless node in the wireless network model has an antenna model that allows traffic flow in certain radius that is modeled as a

110

wireless node with the ability to be able to route them to suitable targets. The wireless nodes are atomic DEVS models connected to each other with wireless network connections.Every wireless node in the wireless network model has an antenna model that allows traffic flow in certain radius that is modeled as a wireless node with the ability to be able to route them to suitable targets. The

115

wireless nodes are atomic DEVS models connected to each other with wireless network connections. 3.1. Atomic node model Wireless nodes are designed as atomic nodes which can generate and route data packets and control packets. Figure 2 presents the wireless node archi5

Page 8 of 27

Figure 1: DEVS-M conceptual model and components

Experimental Frame

MANET private domain model components

Generator

Routing Model

Route

Queue

Packet format

IPAddress

Topography

us

MANET Coupled Model

cr

Transducer

ip t

DEVS-M Model

Topography Atomic Model

Topology Generator

function

message

Bag

relation

content

digraph

atomic

DEVS Kernel

M

Queue

an

Node

port

SimView coordinators

d

entity

Time View

Ac ce p

DEVS

te

DEVS neutral domain model components

Tracking Environment

120

tecture that the most important module is the routing layer. This module is composed of tables that stored shortest route for the data packet. Atomic node consist of energy, topography, processing time, queue, packaging, Node IP, routing.Figure 5 displays UML diagram for a atomic node. The atomic DEVS model for a node is given as following Node=< X, Y, S, s0 , ta, δext , δint , λ > such

125

that

6

Page 9 of 27

DAT A received} Y = {!IIZC send, GIZC send,

ip t

X = {?IIZC receive, DAT A receive,

us

DAT A updated}

cr

DAT A f orward, DAT A send,

S = {(d, σ)|d ∈ {wait, IIZC send, GIZC send, DAT A f orward,

an

DAT A send, DAT A updated}, σ ∈ T ∞}

M

s0 = (DAT A send, 0.01) ta (s) = σ for all s ∈ S δext (((wait, σ), te ), ?DAT A receive) = (DAT A f orward, 0.01)

d

δint (DAT A f orward, σ) = (wait, ∞)

te

δint (wait, σ) = (DAT A f orward, 0.01)

Ac ce p

λ(DAT A f orward, σ) =!DAT A f orward λ(wait, σ) = φ

Energy model that is included in topography model represents the battery

life of the node. Topography model also checks the location and velocity information of the node. All nodes in the experimental framework detect the nodes in the coverage area with the help of the topography model and can connect

130

to these nodes. Processing time is represents the time for the events realized in the node. Queue is the module in which packets that arrive at the node are separated according to types and held aside for processing. Packaging is the module in which packets produced by the node wait to be sent out. Node IP is the module that keeps the node address. Data packets produced by the event

135

generator are received by the node in the event input. Data packets that reach their destination are sent to event generator via event output. Topography in7

Page 10 of 27

Packaging Model

Transport Layer

Routing Model(AntODV)

ip t

Application Layer

cr

Node IP

Packet Model

us

Figure 2: Wireless node architecture model

an

Routing Layer

Processing Time Model

M

Queue Model

Physical Layer

d

Topography Model

MAC Layer

put is used to receive the location, velocity and energy information of the node.

te

After working for a specified processing time, all nodes send their location, velocity and energy information to the topography model from the topography output. Input output allows connection among nodes.

Ac ce p

140

Figure 3 presents the visual appearance of a node in DEVS-M screen. It

includes the ports that allow the dispacth and recieval of packets, node identity, the status of the router and the time for the next event. A simple MAC protocol was designed in the system. MAC protocols controls the data flow between two

145

nodes in the MAC layer. It is situated between the physical layer and the routing layer. MAC protocol used in the study was regarded as ideal. 3.2. Topography atomic model Atomic (topography) model is used for node placement, energy control and mobility. Couplings in mobile topologies need to be updated dynamically. This

150

task is undertaken by the atomic (topography) in order for the simulation envi-

8

Page 11 of 27

Figure 3: Wireless node architecture model

ip t

Node ID Input

Output

Event input

cr

Event output

Topography input Node state

us

Topography output

an

Next event time

ronment to work rapidly. It also allows the system processes to perceive nodes with depleted energy. Figure 4 displays the variables used by the atomic (to-

M

pography) model and its operation system. All nodes is connected with atomic (topography)node over intopo and outtopo port. Atomic (topography)node 155

generate random coordinate for start and end point in a certain area. It gen-

d

erate velocity that is traveled between start and end point. Random velocity

te

varies depending on the vehicle or human .Energy model has constant value that includes a mathematical model. Constant value decreases during the pro-

Ac ce p

cess of transmitting and receiving. When it drops below the threshold value, 160

visualization model change the color of the node. 3.3. MANET Packets

There are two different packets in the network topology: data and control

packets. Data packets are used in routing data to targets and control packets are used for network monitoring and management tasks such as detection of

165

route paths and neighbor identification. Structures of control and data packets are different from one another. Control packets are termed as packer, scouts and

Energyin algorithms (x,y) that use Velocity foragers social creatures such as bees or ants. Developed control packets are modelled as smart agent packets that are active, proactive, able to learn and communicate and able to move. 170

The packets are developed in the network topology in the DEVS model and

(b)

(a) 9

Page 12 of 27

Figure 4: Topography atomic conceptual model

cr

Coordinate

Velocity

an

us

Energy

ip t

Random Generator

Atomic Node

M

Visualisation

they travel in the form of messages. Due to objective structure of JAVA, these

d

packets can be transformed into desired control packets. Packets that arrive at the nodes are placed at the queue based on FIFO logic. The sensor atomic model

175

te

processes the received packet based on its type in line with routing principles. Node queues are limited as in real life. Therefore, when the queue is full, data

Ac ce p

loss may be experienced in the packets. In control packets, when time to live (TTL) (which is defined in line with routing principles) ends; the packet is dropped.

Information that forms the protocol header such as packet identification

180

number, type, TTL, source address and target address is kept in DEVS environment in the variables included in packet objective. Packet objective is prepared compatibly with the real environment in line with the simulation environment. If the route for the data packet to follow is not included in the node, the node

generates a control packet. Data packets are kept in a different queue in the 185

node that will send data and are kept there until a route is identified. Therefore, data packets are transmitted to target without corrupting its content.

10

Page 13 of 27

3.4. Coupled Model

ip t

A coupled MANET model can be generated via DEVS coupled model approach by combining the nodes whose atomic models are formed. Various types 190

of network topologies can be formed with the help of coupled model approach.

cr

Figure 6 presents the connections among nodes in DEVS-M environment. While The coupled model for developed framework consist of generator, transducer, to-

us

pography atomic model, nodes and topology generator, MANET coupled model composed of topography atomic model and nodes. The nodes are connected with 195

inEvent and outEvent ports to generator that generates MANET data packets.

an

Transducer is connected to generator and writes text file events that realize in network. The nodes are connected with NIC-in and NIC-out ports each other. If a node’s transmission area enter other nodes, they can connect with each

200

M

other. These show fine lines in simulation operation area, but don’t show in visualization screen. The coupled DEVS model for a wireless network is given as following

d

N =< X, Y, D, {Mi }, Cxx , Cyx , Cyy , Select >

te

X = {}; Y = {}

D = {node1, node2, ..., nodei} Mnode1 , Mnode2 , ..., Mnodei

Ac ce p

205

Cxx = {}

Cyx = {(node1.!DAT A send, node2.?DAT A receive, ..., nodei.?DAT A receive), (node1.!DAT A receive, node2.?DAT A receive, ..., nodei.?DAT A send)}

Cyy (node1.!DAT A send) = φ,

210

Cyy (node2.!DAT A send) = φ, ..., Cyy (nodei.!DAT A send) = φ. 3.5. Topology generator

Simulation environments develop modeling and simulation tools. Topology generators create network topologies suitable for real world environments with the help of these tools. Therefore, topology generator is an important compo215

nent for simulation tools. A simulation tool without a good generator cannot produce efficient results since it cannot use real world data. 11

Page 14 of 27

BRITE (The Boston University Representative Internet Topology Generator) topology generator is commonly used by researchers in real-like simulation

220

ip t

works and in the investigation of success performance of the network. The fact

that BRITE topology structure is modelled well significantly contributes to the

cr

analysis and development of communication technologies. BRITE topology gen-

erator is highly crucial in network communication, productive protocol designs,

us

solution of problems (management, routing, use of resources etc.), forming the accurate model for simulation and error tolerance studies. In short, the network 225

success mostly depends on BRITE topology generator [14].

an

BRITE topology generator produces network topologies accurately and presents them to the use of researchers who study the accuracy of protocol and algorithms and work on new generation powerful models [15]. An important study

230

M

on topology generators was undertaken by M. A. Rahman et. al. [14]. BRITE topology generator was preferred in the current study since it supports large scale networks, allows for generating topologies easily and is open-source coded.

d

BRITE topology generator was used by adapting it into DEVS-M simula-

te

tion environment. By including BRITE topology generator in the DEVS-M environment, it is possible to generate desired size MANETs. BRITE topology generator includes options such as MANET area desired to be generated,

Ac ce p

235

number of nodes, network model and node placement. It is possible to fully intervene in the MANET model with the help of these characteristics. 3.6. Visual tracking

A graphic interface was developed to monitor the locations and movements

240

of generated network topologies in real environment. Therefore, it was possible to observe the connections between random node placements, distribution areas, coverage areas and connections to one another.The distance between two pixels in DEVS-M visualization screen was accepted to be 1m. Figure 7 presents the screen shot of a topology formed by 10 nodes in a 500 X 500 area. Developed

245

visualization screen shows the adaptation of MANET degenerated in DEVS-M environment to the real world. As in real life, Figure 7 shows that 10 nodes are 12

Page 15 of 27

distributed in a 500 X 500 area. Small circles represent the nodes. Big circles show the coverage area.The changes in DEVS-M environment can be accurately

4. Simulation experiments

cr

250

ip t

visualized in the visualization environment.

Scenarios produced under different number of nodes, different number of

us

data packets and different network deployment, obtained data and the observed outputs were offered in this study. AntODV routing protocol was used in the

255

scheme developed from AODV protocol.

an

scenarios. AntODV protocol is an ant colony optimization based load balancing

Table 1: The simulation parameters in experiments

DEVS-M

M

Simulation Parameters

AntODV

IP Address model

IPv4

IP Address size (byte)

4

Message type

Modified IP Packet

te

d

Routing protocol

552

Scale

100 nodes

Ac ce p

MTU (byte)

Simulation time (sec)

300

Simulation area

500X500 m

Node Transmission Range

250 m

Node Velocity

1-10 m/sec

Node Deployment Model

Waxman

Operations were done via computer using the confirmed parameters from

simulation experiments explained in Table 1. simulation results were tested with JRE 1.7 on computers with Intel Core i7, 2.0 GHz, 4GB RAM, 64 bit Ubuntu. During the simulation process, parameters of the simulation environ260

ment, obtained data (statistics, average network output, packet loss, etc.) and information regarding the transmitted message in the network were interpreted 13

Page 16 of 27

by event transducer and saved in the file (output.csv) with file name extension

ip t

CVS (comma separated values) and were presented with graphics.

5. Simulations results and discussion

In the first simulation, comparisons were done for throughput and average

cr

265

latency of the packets with 60, 70, 80, 90 and 100 nodes. In the second sim-

us

ulation, packet numbers were changed to 800, 900, 1000 packets and similar comparisons were maintained.

270

an

5.1. Throughput

Figure 8 and Figure 11 presents the throughput of the routing algorithm used in the simulations undertaken in the framework of the study. Figures show

M

that throughput was quite little unaffected from these changes although the number of nodes change in framework. In Figure 8 and Figure 11, the throughput that decrease somewhat toward simulation time end is occurred by latency. Besides, fluctuations are occurred by loss packet. Fluctuations and decreased

d

275

te

throughput are observed because MANETs composed of mobile nodes. In Figures, when the number of packets is changed in framework, throughput still

Ac ce p

has largely maintained its stability. The developed simulation kept its stability even when it was undertaken in different topologies, with different packet num-

280

bers and with different numbers of nodes. It is observed that throughput can be maintained stability in algorithms developed by using ant colony algorithms [16]. AntODV protocol which is an ant colony optimization based load balancing scheme derived from AODV protocol provided the expected results. This finding shows that the developed simulator is successful in simulating MANET.

285

5.2. Average latency Average latency figures are seen in Figure 10 and Figure 11 . Figures show that increase in the number of nodes or the number of packets resulted in the expected outcome and latency linearly increased. Increase in the traffic density 14

Page 17 of 27

290

in the system which was an expected result also increased the latency. In the study, use of optimization in the routing algorithm decreased the amount of

ip t

the increase. This finding shows that MANET simulator is close to obtaining

the expected results in reality. The consistency mentioned about throughput is

295

cr

also observed for average latency. The current finding points to the fact that it is possible to obtain results compatible with what is expected via DEVS-M

us

simulator developed for MANET. 5.3. Memory Consumption

an

Memory consumption of the simulator developed for MANET was tested. Table 2 presents memory consumption for per connection [17]. Compared with 300

the tests for different simulator tools in the literature, memory consumption for

M

per connection was found to be lower than ns-2 simulator as seen in Table 2. Figure 12 and Figure 13 presents memory consumption for per connection and memory consumption for all connections. Examination of figures show

305

d

that memory consumption for all connections is almost constant while memory consumption per connection decrease . This finding shows that increasing the

te

number of nodes does not create extra load for the system. Therefore, the developed simulator did not lose any of its velocity although the numbers of

Ac ce p

nodes increased. In other simulators, used memory consumption increases along with the increase in network size [18]. This slows down the system due to

310

additional load brought by memory. Table 2: Comparison of memory consumption of network simulation tools

Simulator

Memory consumption for per connection

ns-2

93,3 KB

J-Sim

21,7 KB

Java-SSFNet

53,3 KB

C++-SSFNet

78,1 KB

DEVS-M

72,5 KB

The tests undertaken in the study show that although the developed MANET 15

Page 18 of 27

simulator is better than only ns-2 simulator in terms of memory consumption for per connection, it seems better than other simulators in terms of the extra

than other simulators in terms of scalability.

cr

315

ip t

loads that all connections bring to the network. Therefore, DEVS-M is better

6. Conclusion

us

Developed simulation framework was developed by adapting DEVS-Suite simulator tool using DEVS formalism. AntODV protocol based on ant colony optimization, being load balancing scheme, derived from AODV protocol was used to test the developed simulation framework accuracy and to present its

an

320

differences from the other simulation tools. Average latency tests show that

M

latency was due to node density or traffic intensity. DEVS-M simulator provides close findings to expected results in the real life. Amount of increase in latency was decreased by AntODV load balancing scheme and so throughput is able to stability. DEVS-M simulator was found to produce lower results than ns-

d

325

2 simulator in terms of memory consumption for per connection but higher

te

than other simulators. The simulator kept its stability in terms of memory consumption for all connections and the increase in the number of nodes did

Ac ce p

not affect memory consumption. This fact also shows the superiority of the

330

simulator compared to the others. Since network size does not place extra load on memory consumption, the developed simulator may be transferred to web environment by using the fact that JAVA is independent of platform. A platform-free web based simulator training environment may be designed by using the innovations presented by DEVS-M simulator.

335

References

[1] S. Kurkowski, T. Camp, M. Colagrosso, Manet simulation studies: The current state and new simulation tools, Mobile Computing and Communications Review 9 (4) (2005) 50–61.

16

Page 19 of 27

[2] T. Issariyakul, E. Hossain, Introduction to network simulator NS2, Springer Science & Business Media, 2011.

ip t

340

[3] X. Chang, Network simulations with opnet, in: Proceedings of the 31st con-

ference on Winter simulation: Simulation—a bridge to the future-Volume

cr

1, ACM, 1999, pp. 307–314.

[4] B. P. Zeigler, H. Praehofer, T. G. Kim, Theory of modeling and simulation: integrating discrete event and continuous complex dynamic systems,

us

345

Academic press, 2000.

an

[5] A. Bourouis, B. Belattar, Eqnm2l: Towards a metamodel for extended queuing networks, International Journal of Computer Science Issues

350

M

(IJCSI) 9 (1).

[6] S. Kim, H. S. Sarjoughian, V. Elamvazhuthi, Devs-suite: a simulator supporting visual experimentation design and behavior monitoring, in: Pro-

d

ceedings of the 2009 Spring Simulation Multiconference, Society for Com-

te

puter Simulation International, 2009, p. 161. [7] F. C ¸ elik, A. Zengin, S. Tuncel, A survey on swarm intelligence based routing protocols in wireless sensor networks.

Ac ce p

355

[8] F. C ¸ elik, Aco-based load balancing scheme for manets, Technical Gazattedoi:10.17559/TV-20140715145318. URL http://dx.doi.org/10.17559/TV-20140715145318

[9] C. E. Perkins, E. M. Royer, Ad-hoc on-demand distance vector routing,

360

in: Mobile Computing Systems and Applications, 1999. Proceedings. WMCSA’99. Second IEEE Workshop on, IEEE, 1999, pp. 90–100.

˘ [10] F. Celik, A. Zengin, B. C ¸ OBANOGLU, Discrete event simulation-based performance evaluation of internet routing protocols, Turkish Journal of Electrical Engineering & Computer Sciences 21 (6) (2013) 1720–1736.

17

Page 20 of 27

365

[11] B. Cobanoglu, A. Zengin, H. Ekiz, F. Celik, A. Kiraz, F. Kayaalp, Implementation of devs based distributed network simulator for large-scale

ip t

networks, International Journal of Simulation Modelling (IJSIMM) 13 (2).

[12] B. C ¸ obano˘ glu, A. Zengin, S. T¨ uncel, H. Ekiz, A survey of network modeling

370

cr

and simulation tools: Devs comparison.

[13] A. C. Chow, Parallel devs: A parallel, hierarchical, modular modeling for-

Computer Simulation 13 (2) (1996) 55–68.

us

malism and its distributed simulator, TRANSACTIONS of the Society for

an

[14] M. A. Rahman, A. Pakstas, F. Z. Wang, Network topology generation and discovery tools, Proc. of the 7th EPSRC Annual Postgraduate Sym-posium 375

on the Convergence of Telecommunications, Net-working and Broadcasting

M

(EPSRC PGNet 2006) (2006) 26–27.

[15] A. Medina, A. Lakhina, I. Matta, J. Byers, Brite: An approach to universal

d

topology generation, in: Modeling, Analysis and Simulation of Computer

380

te

and Telecommunication Systems, 2001. Proceedings. Ninth International Symposium on, IEEE, 2001, pp. 346–353.

Ac ce p

[16] G. Di Caro, F. Ducatelle, L. M. Gambardella, Anthocnet: an ant-based hybrid routing algorithm for mobile ad hoc networks, in: Parallel Problem Solving from Nature-PPSN VIII, Springer, 2004, pp. 461–470.

[17] D. M. Nicol, Scalability of network simulators revisited, in: Proceedings

385

of the Communication Networks and Distributed Systems Modeling and Simulation Conference, 2003, pp. 1–8.

[18] E. Weingartner, H. Vom Lehn, K. Wehrle, A performance comparison of recent network simulators, in: Communications, 2009. ICC’09. IEEE International Conference on, IEEE, 2009, pp. 1–5.

18

Page 21 of 27

Ac ce p

te

d

M

an

us

cr

ip t

Figure 5: UML diagram of a node

19

Page 22 of 27

ip t cr

Ac ce p

te

d

M

an

us

Figure 6: DEVS-M coupled model

20

Page 23 of 27

M

an

us

cr

ip t

Figure 7: DEVS-M visulasition area(a)when simulation starts(b)when simulation ends

(b)

d

(a)

70 Nodes

80 Nodes

90 Nodes

100 Nodes

Ac ce p

Throughput

60 Nodes

te

Figure 8: In different scales DEVS-M throughput according to number of nodes

0,2 0,18 0,16 0,14 0,12 0,1 0,08 0,06 0,04 0,02 0

Simulation time(sec)

21

Page 24 of 27

Figure 9: In different scales DEVS-M throughput according to number of packets

900 Packets

1000 Packets

ip t

800 Packets

0,35 0,3

cr

0,2 0,15

us

Throughput

0,25

0,1

an

0,05 0

M

Simulation time(sec)

Figure 10: In different scales DEVS-M average latency according to number of nodes

0,16

90 Nodes

100 Nodes

Ac ce p

Avarage latency (sec)

0,14

80 Nodes

d

0,18

70 Nodes

te

60 Nodes

0,12

0,1

0,08 0,06 0,04 0,02

0

Simulation time(sec)

22

Page 25 of 27

800 Packets

900 Packets

1000 Packets

cr

0,1

us

0,08 0,06 0,04 0,02

an

Avarage latency (sec)

0,12

ip t

Figure 11: In different scales DEVS-M average latency according to number of packets

M

0

d

Simulation time(sec)

te

Figure 12: DEVS-M memory consumption per connection

95 90 85 80 75 70 65 60 55 50 45 40

Ac ce p

Memory Consumption (KB)

Per connection

60

70

80

90

100

Number of nodes

23

Page 26 of 27

ip t cr us

Figure 13: DEVS-M memory consumption per connection

an

All connection

M

6000 5000

d

4000

te

3000 2000 1000

Ac ce p

Memory consumption (KB)

7000

0

60

70

80

90

100

Number of nodes

24

Page 27 of 27