Home

The Madhoc Metropolitan Adhoc Network Simulator

1. oe Bu ho 7 s gt eo r e 8 Figure 3 2 4 000 nodes moving according to the market place environment 3 2 1 On metropolitan mobility Simulating metropolitan mobility is difficult because urban zones feature an extensive list of dissimilar configurations avenues pedestrian areas places shopping malls etc Building a generic model that take into consideration all the components of a city is a daunting task So far attempts at modeling metropolitan mobility focus on specific configurations The Manhattan mobil ity model is the most relevant initiative The city chapter and the graph based mobility models 9 are generalizations of it In the Manhattan model the sta tion is allowed to move along the horizontal or vertical streets on the urban map Unfortunately the model fails at representing important characteristics of metropolitan environment such as the existence of shopping zones the width of the streets etc Moreover nothing is said about the propagation of radio waves 3 2 2 Human mobility a mobility model driven by the intentions of humans Humans do not move randomly When moving they have a determined target spot and move towards it The target may be a few meters away next shop next crossroads other sidewalk etc as well as far away next district next city etc 3 2 HUMAN MOBILITY 23 Market place Probability in 0 1 Figure 3 3 On the left an example of a
2. 19 C Kunze U Grossmann W Storka and KD Muller Glaser Applica tion of ubiquitous computing in personal health monitoring systems In DGBMT Jahrestagung der Deutschen Gesellschaft F r Biomedizinische Technik pages 360 362 2002 20 Johannes K Lehnert Daniel G rgen Hannes Frey and Peter Sturm A scalable workbench for implementing and evaluating distributed applica tions in mobile ad hoc networks In WMC 04 Western Simulation Multi Conference pages 154 161 2004 BIBLIOGRAPHY 61 21 William Navidi and Tracy Camp Stationary distributions for the ran dom waypoint mobility model IEEE Transactions on Mobile Computing 3 1 99 108 2004 22 Guangyu Pei Mario Gerla Xiaoyan Hong and Ching Chuan Chiang A wireless hierarchical routing protocol with group mobility n WCNC1999 IEEE Wireless Communications and Networking Conference number 1 pages 1538 1542 IEEE IEEE September 1999 23 Ray and Suprio Realistic mobility for manet simulation December 2004 24 Sebastien Matas Riera Oliver Wellnitz and Lars Wolf A zone based gam ing architecture for ad hoc networks In NETGAMES 03 Proceedings of the 2nd workshop on Network and system support for games pages 72 76 ACM Press 2003 25 Hartmut Ritter Min Tian Thiemo Voigt and Jochen H Schiller A highly flexible testbed for studies of ad hoc network behaviour In LCN pages 746 752 2003 26 E Royer and C Toh A review of current r
3. 44 CHAPTER 6 ADVANCED USAGE void configure void doIt double time String getName requires that you simply define and return a name for this application The name identifies the application so take care at not using a name that is already in use public String getName return hello message application The void configure method codes how the application will handle its configuration parameters In our present case the configuration simply indicates to which frequency the application will emit hello messages public void configure throws Throwable this helloFrequency 1000 x getMonitor getNetwork getSimulation getConfiguration getInteger hello application frequency Obviously the helloFrequency field must have been defined In order this to work the hello_application_frequency configuration key must be defined in any configuration file Here it is advisable to create a new configuration file that will be dedicated to our new application The void doIt double time method is invoked at each iteration of the simulator This is where what the application do is actually coded An application that periodically sends its identity to its neighbors could be implemented in the following way public void dolt double time if getSimulatedTime this helloFrequency 0 Message message new Message TransferableDouble id new TransferableDouble id setValue getComput
4. Mobile phones PDAs and laptops do not feature for the moment the same communicating technologies More precisely the very limited energy storage ofa mobile phone refrains it from using powerful radio signals Hence mobile phones 7 8 CHAPTER 2 MODELS MessageBuffer ncomingMessageQueue O NetworkingUnit NetworkingTechnology point networkingUnit networkingUnit Dn 0 Q 7 Networkinterface Figure 2 1 The station model networkType as well as PDAs are generally operating in a 8 12 meters range thanks to their built in Bluetooth adapter although some mobile phone PDAs equipped with WiFi interface are being appearing for a short while On the other hand all today s laptops feature WiFi network interfaces all new devices come with a high speed WiFi interface which provide a bandwidth up to 54Mbps Some of them also integrate a Bluetooth adapter These divergences in the way devices are equipped have a serious impact on the topology of the network Imagine a Bluetooth mobile phone moving in an area populated with dozens of WiFi laptops Since WiFi and Bluetooth are incompliant the mobile phone won t be able to establish any connection with its geographical neighbors madhoc currently supports WiFi IEEE802 11b Bluetooth and Wireless USB These protocols which mostly stand at the PHY and MAC layers are not modeled with detail Protocols are represented in terms of e bandwidth the bandwidth is shared by
5. 13 D Gorgen H Frey and C Hutter Information dissemination based on the en passant communication pattern KiVS Fachtagung Kommunikation in Verteilten Systemen 2005 14 Xiaoyan Hong Mario Gerla Guangyu Pei and Ching Chuan Chiang A group mobility model for ad hoc wireless networks In MSWiM 99 Pro ceedings of the 2nd ACM international workshop on Modeling analysis and simulation of wireless and mobile systems pages 53 60 ACM Press 1999 15 EPFL Information Sciences Institute The nab network in a box wireless network simulator In http nab epfl ch 2004 16 Amit Jardosh Elizabeth M Belding Royer Kevin C Almeroth and Sub hash Suri Real world environment models for mobile ad hoc networks Journal on Special Areas in Communications Special Issue on Wireless Ad hoc Networks 14 2 January 2005 17 Amit Jardosh Elizabeth M Belding Royer Kevin C Almeroth and Sub hash Suri Towards realistic mobility models for mobile ad hoc networks In MobiCom 08 Proceedings of the 9th annual international conference on Mobile computing and networking pages 217 229 ACM Press 2003 18 Per Johansson Tony Larsson Nicklas Hedman Bartosz Mielczarek and Mikael Degermark Scenario based performance analysis of routing pro tocols for mobile ad hoc networks In MobiCom 99 Proceedings of the 5th annual ACM IEEE international conference on Mobile computing and networking pages 195 206 ACM Press 1999
6. And so on This research is on progress Chapter 9 OOP implementation in java The madhoc project has begun in January 2004 Since then 20 000 lines of code have been written The code organized in 205 classes classified in 42 packages All along the development process a great attention has been paid to the elegancy of the design and to the quality of the algorithms The resulting code is clean and stable madhoc simulations deal with numerous entities of the same class the sta tions Such architecture is used many other software like in multi agent frame works Experience has shown that modeling the entities as columns in a huge array of primitive objects could improve the simulations efficiency by factor 10 However doing this merely discards the advantages of OO programming mod ularity reusability But we initially considered that using such a technique would have prevented us from developing madhoc as it is now and would greatly compromise its open model 55 56 CHAPTER 9 OOP IMPLEMENTATION IN JAVA Chapter 10 Grid madhoc What constitutes the main obstacle to the constitution of large grid is general network security Generally the client server defines that only one daemon server software of a given type runs on a given computer Things do not work like this in madhoc s grid It defines that exactly one server called simulation server runs on every processor Then a workstation equiped with n
7. the radio signal Shops are randomly located within the mall so as the distance between two given shops must not be lower than 10m People within a given shop move according to the random waypoint mobility model at a speed chosen in between 1 and 4 km h Pauses last between 1 and 20 seconds At most 20 pauses are allowed When a station goes out of a shop it randomly choses its next destination shop Second a corridor is a bi directional path connecting two shops There is always one corridor connecting two given shops People walk in corridors at a speed randomly chosen between 2 and 6 km h In order to model the broadness 53 54 CHAPTER 8 TARGETTED APPLICATIONS of the corridors a variation of the direction of people walking into them is tolerated The broader is the corridor the bigger variation is tolerated The mall mobility model generates the emergence of dense and highly con nected regions These regions are sometimes isolated At runtime one can observe that they get sporadically get connected to other regions thin paths like chains This kind of topology illustrates that applications as well as protocols must deal with high variations of the density Pedestrian area mobility The pedestrian area mobility model simulates the mobility of people walking in a pedestrian only city area This mobility model is an extension of the graph based mobility model 28 More precisely the street network is represented as a grid eac
8. 19190089727 20 J 150005727 El 10 150863727 75 17 i y 013306001 sot 0015305001 501 hosssusunn u 3306004 Figure 5 4 A view that give an overview of all the numerical measures available in the simulation process 2D representation view The 2D representation view aim to improve the functionality of the numerical measure list view It permits the user to select two measures on the latter view 40 CHAPTER 5 THE GRAPHICAL USER INTERFACE and graphical expose the relations between them on a 2D axis system The way the measure have to be selected is intuitie The control panel on the right side of the 2D representation view shows two checkbox that indicate which dimension is currently being configured By default the X dimension is selected as shown on image 5 5 Then click on whatever measure within the numerical measure list view It can observed that the X axis of the 2D representation is immediately updated with the name and color of the selected measure Now change the selected dimension using the appropriate checkbox on the 2D representation view s control pane Then click on some other measure in the numerical measure list view Just like what happened for the X axis the Y one is updated madhoc now shows a color 2D mathematic function which highlight the relation betweent the two selected measures as shown in the example image 5 6 Most of the time the user will
9. Altering the resolution of the simulation 8 Targetted applications 8 0 4 Broadcasting 8 0 5 Mobility models 8 0 6 Ad hoc computing towards the ad hoc grid 9 OOP implementation in java 10 Grid 43 43 43 46 46 46 46 48 49 51 51 51 53 53 53 54 55 57 Chapter 1 Introduction Note This document stands for a user manual and a technical reference for the mad hoc simulator Writing such a document is difficult because several approaches are possible in defining its structure and there do not exist any precise guidelines of how to proceed In this document we will try to keep things as simple and logical as possible Obviously we cannot give all details of madhoc in one single document Hence in addition to this pages the reader should browse the source code which is publicly available Note that if the source code is not completely documented its structure should be easily understandable to people used to the OO programming paradigm and to the Java language Mobile ad hoc networks MANETs are networks composed of a set of com municating devices called nodes all along this article able to spontaneously interconnect without any pre existing infrastructure Stations in range to one another communicate in a point to point fashion In addition to that these nodes are generally mobile The main interest in MANETs is growing Not on
10. When the GUI starts it opens a desktop environment in which every win dows correspond to a distinct projection 2 3 3 of the simulated network as illus trated on image 5 1 Projections are easily identifiable since their name entitles the corresponding window aver i oran cime Figure 5 1 On start up madhoc open one window for each projection available All windows have the same features A window is split vertically in two zones As shown on image 5 2 the left side of the window shows a tree widget graphical component This tree is three levels deep and is organized as follows 37 38 CHAPTER 5 THE GRAPHICAL USER INTERFACE root is the name of the simulation first level shows the monitors 2 2 2 available second level the deepest one shows the views made available by each monitor see section 5 1 whole network A P Whole network amp broadcasting DFCN a 7 EH network_monitor Aircraft view 9 random_waypoint_mobility Random waypoint view E simulation_monitor se distribution plotter E 2D plotter Numerical measures D Figure 5 2 A tree for selecting the views 5 1 Monitor views A view is a widget that permit the graphical rendering of some information exposed by its monitor Views are shown on the right side of the projection window More than one view is showable simultaneously Showing one single views is achievable by si
11. all communicating devices operat ing on a common media All devices have the same chance to send receive data e range of coverage define the maximum distance to from which the devices can receive data e packet size emitted data is organized into packets A packet is the small est chunk of data that can be emitted over the network for a given pro tocol e data transfer cost see figure2 2 define the price for emitting one byte over the network madhoc defines several basic cost models This model permits to consider the impact on communication technologies on network connectivity and bring heterogeneity 2 1 NETWORK MODEL 9 Figure 2 2 madhoc features a cost model Different computational storage capacities Initially targeting the analysis of the future ad hoc grids madhoc integrate a basic model for the ad hoc grid node More precisely it endows the nodes with computational power and storage ability madhoc assumes that those abilities depends on the type of device considered A laptop will generally embed a fast processor coupled with a quite large central memory and a several gigabytes hard disk PDAs and mobile phone embeds energy efficient components hence their computation capacities is severely limite
12. both broadcasting and mobility 2 3 3 Simulation area and projections Simulation area For the sake of simplicity madhoc defines that nodes evolve in a square shape surface The surface of the simulation area can be defined by using the simulation_area_sur face It is expressed in square meters Because madhoc relies on a specific technique called bining bining is a general modeling technique which consists in dividing a considered zone in a grid multi dimensional griding is possible so as oper ations on the zone are executed with a lower complexity the surface exhibits the following constraint it must be multiple of 10 000 The number of nodes in the simulation area is constant nodes cannot step out of the simulation area as well as no new node can step in Projections In many cases the user will want to consider only a sub network For example the user may want to observe what is going on only on a specific zone of the sim ulation area he may want to observe what happens only on Bluetooth nodes etc In order to make this possible madhoc defines the concept of network projection Just like the mathematical definition of projection a projection of the simulated network will result in a set of the nodes that match the crite ria defined by the projection itself The communication links retained in the projection are the links whose two participants are part of the projection The sensors available in the simulation will be
13. by the UP project 2 42 CHAPTER 5 THE GRAPHICAL USER INTERFACE Chapter 6 Advanced usage 6 1 Creating a new module Section gives the initialization process of a simulated application This helps at understanding what needs to be written by the programmer at the development stage What need to be hardcoded in Java are the monitor and the code of the application These two components are enough for deploying and executing the application It is very likely that the user will want to have some feedback on the application he is developing Then he needs to write one or more sensors that will periodically take measures on the application and report them to the user interface be it graphical or console based Optionally if the user wants to use the advanced features of the graphical user interface he has the possibility to define new views for his applications These views which are Swing components will be dynamically integrated in the madhoc s GUI 6 1 1 Example The best way to know how to create a new application for madhocis to look at the following example In the following we will define a simple application that merely periodically sends hello messages to its neighbors The code of the application First you need to define what your application does This is done by deriving the class org lucci madhoc network Application To do this you need to implements the following methods String getName 43
14. code it is a data structure that has nothing to do with the configuration keys available However in the framework mode the configuration object must be initialized with the set of keys provided by the user Decoupling the keys from the data structure that is actually used by the simulation may sound strange Indeed it is strange but it has the great ad vantage that the keys available actually the keys allowed are hardly coded in Java This way the user cannot use non existing keys More over programmers who use a Java editor which provide code completion can use the latter for completing the name of the keys The configuration data structure must then be initialized in the following way Configuration config new Configuration config load keys Then a simulation object must be created and initialized with the configura tion The configuration contains enough data to allow a complete initialization of the simulator In case some configuration keys are missing or have been fed with unsuitable values explicit errors will be shows as java exceptions Simulation simulation new Simulation simulation configure config At this step the simulation is completely initialized and is hence ready to use execute one single iteration of the simulator simulation iterate madhoc does not run until the simulation has completed The simulation must be progressively advanced by the user A typical way of executing a simulation is
15. file not a directory whose the name ends with madhoc it is considered to be a configuration file and is used as such If the given command line parameter specifies a directory the latter is scanned to configuration files A full example of a configuration split in several files for the sake of clarity can be downloaded at http www lih univ lehavre fr hogie madhoc code config zip There exists a long list of configuration keys available in the default madhoc The example configuration file whose the url was given in the previous paragraph contains all the configuration keys allowed in madhoc as it comes by default We did not wish to give long explanation on al configuration keys herein The example configuration file is wealthly documented Please refer to it for further explanations 4 5 USING MADHOC AS A STANDALONE APPLICATION 35 4 4 2 Configuring madhoc as a framework When using madhoc as a framework there is no need to handle text files Here you need to instantiate a org lucci madhoc config ConfigurationKeys object and set its fields to the appropriate values The name of the fields of the configuration object The following example is the same of the one in the previous section adapted to a framework mode ConfigurationKeys config new Configurationkeys config simulation _name my simulation config simulation _resolution 10 The ConfigurationKeys object features all the configuration keys available as publi
16. instance in the real world A monitor aims at maintaining a global view on all applications of a given class and at carrying out operations applicable to a set of applications such as their deployment their initialization and the observation of the decentralized process Measurement The greatest benefit of monitors is that they provide the user with virtual sensors which allow the user to observe what is going on in the simulation process Just like a physical sensor thermometer barometer etc a madhoc sensor will take measures on a given system Measures are taken at each iteration Each new measure is appended to an history There exist several sorts of measures as the phenomenon that are sensed are of various nature If the phenomenon can be sensed as a numerical value like the network throughput and the simulation time that can be expressed as numbers the measure is said to be numerical Tt is important to distinguish numerical measure from the other as their nature allow specific things to be done with them e g mathematical operations graphical rendering etc Unfortunately sensing takes time The more sensors are defined the longer will take the simulation Then for the development process it is advisable to use as many sensors are available in order to detect the slightest flaw of the code but for the research experimentations the user should enable only the metrics he is interested in by discarding the others 2 3 Simul
17. the low bandwidth communication links only a few nodes were able to transmit before the link is saturated the others if any were blocked Moreover these privileged few nodes were always the same the scheduling was hence unfair In order to give all application the same chance to execute and to emit data on the network during an iteration of the simulator the application scheduler executes in a stochastic fashion More precisely along one single iteration all applications are asked to execute multiples times each time for a very short period The order the applications is randomized At each turn all applications are invoked but in an non predefined order 2 3 6 Termination condition madhoc is a general simulator As such it provides a general model for the termination condition of the simulations as explained in this section A simu lation is said to be terminated if all the applications it executes are terminated if the concept of termination makes sense More precisely a broadcasting ap plication can be said to be terminated when all the nodes are reached or when the broadcast message has expired Whatever it is the termination condition is easy to define In the case of beaconing applications the application that steadily sends hello messages in the neighborhood the concept of termina tion does not make sense A beaconing application should never terminate unless the node may want to become invisible whi
18. Pragmatically speaking the way the configuration information is passed to madhoc depends on the way the simulator is used 4 4 1 Configuring madhoc as a standalone application In the case of running madhoc as a standalone application the configuration information must be passed either by the standard input of the process or by providing it as an ASCII text file whose name is specified on madhoc s com mand line Except from the lines beginning with the character which are considered as comment lines and blank lines which all are skipped all lines con stituting a configuration text must strictly respect the following BNF Bacus Normal Form definition line key value value xj key 7 value Here follows as example of correct configuration entries some comment simulation_name my simulation simulation_resolution 10 some other comment Each line is called a configuration entry Entries are mandatory as soon as madhoc ask for them since madhoc does not make any assumption of default values If a configuration entry cannot be found in the configuration provided by the user madhoc immediately interrupts and prints an explicit error message on the standard error The configuration can be wholly contained in a single file as well as it can be split in several files Indeed when scanning its command line parameters madhoc check if a given parameter is a directory or not If the parameter is a normal
19. SZ Nx gt Ps de N a Y x N et y gt y Figure 3 6 On the left an example of a Mall 15 Degree number of nodes network formed by people shopping in a mall On the right the average degree distribution for such networks 3 3 3 Streets The street environment model the network formed by vehicle driving in a city section Here spots represent the crossroads The resulting connectivity exhibit some very high dense regions in terms of communication links which are the crossroads connected by thin paths formed by nodes moving along streets 3 3 REAL WORLD MOBILITY BASED ON HUMAN MOBILITY Parameter name Value Unit Simulation area surface 1000 x 1000 m Spot density 50 place km Spot radius 3 15 m Wall obstruction ratio 0 9 Mobile phone density 5000 node km PDA density 1000 node km Laptop density 500 node km Hotspot density 0 node km Out spot velocity 1 20 m s In spot velocity 0 3 20 m s 0 01 0 005 E Degree number of nodes 27 Figure 3 7 On the left an example of a network formed by cars driving in a city section On the right the average degree distribution for such networks 3 3 4 Supermarket The spot represent the edges of the Parameter name Value Unit Simulation area surface 50 x 50 m Spot density 100 000 place km Spot radius 3 15 m Wall obstruction ratio 0 2 Mobile phone dens
20. The Madhoc Metropolitan Adhoc Network Simulator Luc Hogie Fr d ric Guinand and Pascal Bouvry Universit du Luxembourg Campus Kirchberg 6 rue R Coudenhove Kalergi L 1359 Luxembourg luc hogie pascal bouvry Cuni lu 2Universit du Havre Laboratoire d Informatique 25 rue Philippe Lebon 76600 Le Havre France frederic guinand univ lehavre fr March 22 2006 Contents 1 Introduction 2 Models 2 1 Network model 2 44 so maet e a eee ee 21 1 Hardware raan O Dn nei A ae ete 212 Software 2 ais DER A A 2 2 Ad hoc application model 22 1 Application ata Dres 2 242 Monitor 5 4 0 6 6 Bo A de ia 2 3 Simulation model 2321 Languages aai 4 er ae a ane cn Ne EUR EE ALES 2 3 2 Discrete event discrete time resolution 2 3 3 Simulation area and projections 2 3 4 Random number generation 23 0 Scheduling ira ita un i ee ae nn sua e dnd 2 3 6 Termination condition 2 3 7 Application deployment 3 Scenarios 3 1 Random waypoint 3 11 City places san e ses 48 a nea de de LS 3 1 2 Market place 3 1 3 Concert place 38 022 22 2220 ae en 3 2 Human mobility wm A Pe 4 3 2 1 On metropolitan mobility 3 2 2 Human mobility a mobility model driven by the inten tions Of humans tipon p e
21. a invoke the iterate method in the body of a while loop whose the termination condition consequently the termination condition of the simulation depends on the simulated application The following code simulates one second of network activity iterate until the simulated time is lower than 1 second while simulation getSimulatedTime lt 1 simulation iterate 48 CHAPTER 6 ADVANCED USAGE When the simulation is being running in the body of the while loop or when it has completed after the block it is possible to get some information on the simulation process Indeed the framework features an extensive list of utility methods like getSimulatedTime that can be used for retrieving informations on the simulated network and applications In order to obtain some information that is specific to a given application it is advisable to get the instance of the latter application which is exposed within the simulation as a Monitor object gets the monitor that is in charge of observing what s going on in the network Monitor networkMonitor simulation getNetwork getMonitorMap NetworkMonitor class Sensors are designed for application specific purposes They are hence ex posed by monitors Obtaining a sensor can be done using the following code retrives the sensor that is in charge of measuring the network throughpu Sensor throughputSensor networkMonitor getSensorMap get ThroughputSens As explaine
22. a new extension consists of some class files that contain the executable code and optional configuration file that enables the user to interact with the extension 2 3 2 Discrete event discrete time resolution Unlike many other simulators madhoc does not rely on the discrete event sim ulation paradigm Instead of jumping from event to event madhoc s kernel iterates upon time Time is discrete The time interval between two iterations is called the resolution 14 CHAPTER 2 MODELS The resolution of a simulation process is defined by the user The configu ration key simulation resolution allows the user to modify it You should of course use the same resolution for all the experiments otherwise your result can not be assumed to be consistent The greatest is the resolution the faster and the less accurate is the simulation process The resolution of the simulation ideally depends on the application you want to simulate In the specific case of RAD based broadcasting the resolution should be at least twice lower than the maximum RAD otherwise the benefit of using a RAD is simply lost In the spe cific case of the simulation of mobility the resolution should ensure that mobile node move with reasonably small steps otherwise some connections that would have occurred in the real world would not be simulated In our case RAD based broadcasting in mobile network the definition of the resolution should fit the constraints brought by
23. according to the random mobility model The speed of nodes moving within a spot is not defined by the human mobility model as it depends on the kind of environment that is modeled By randomly moving within a given place nodes soon or later reach the edge of the place then they are considered to be gone out of the place At this moment the next target place it chosen The decision of the next target place is based on the list of places that still have not yet visited and on the distance 24 CHAPTER 3 SCENARIOS Probabilty in 0 1 Figure 3 4 On the left an example of a network formed by people attending a concert On the right the average degree distribution for such networks between the current location of the node and the targets so as the closest places are preferred Once the list of places that have not yet visited is empty it is refilled with all the places in the simulation Emerging topology of human mobility As shown on figure 3 6 the human mobility model can generates the emergence of dense and highly connected regions These regions are sometimes isolated meaning that they are not connected to the rest of the network At runtime one can observe that they sporadically get connected to other regions thin paths like chains This kind of topology illustrates that applications as well as protocols must deal with high variations of the density By changing the parameters of the human mobility we can achiev
24. adhoc is a metropolitan ad hoc network simulator targeting the investigation of ad hoc grid computing 3 It features the components required for both realistic and large scale simula tions as well as the tools essential to an effective monitoring of the simulated applications Chapter 2 Models In this section we will present sequentially the models for the network for the ad hoc applications and for the simulation engine 2 1 Network model A good place to start with madhoc is probably to understand what does it simulate In this section we distinguish the hardware and software parts of the network model 2 1 1 Hardware madhoc models an heterogeneous mobile ad hoc network Mobility does not mean that all nodes are in a continuous move It simply defines that nodes have the ability to move Different devices On today s electronics market mobile communicating devices can be found em bedded on laptops PDAs and mobile phones These machines are very different in their technical aspects and in the way they are used More precisely Laptops are much more capable in terms of computations than a PDA and the latter is in turn much more capable than a mobile phone A mobile phone is mobile as well as PDA and laptops However when switched on PDAs and laptops usually do not move except in the case of people working in a vehicle This constitutes the major difference between these devices Different networking technologies
25. aphical interface madhoc allows the user to create graphical views on the applications it develops This views which are standard Swing components will be dynamically integrated into the standard interface Moreover their content will be updated at each iteration of the simulation process The user can then have a precise view of the simulated application 18 CHAPTER 2 MODELS Chapter 3 Scenarlos In 6 Bohacek states that a simulator does not always need to be realistic but only must stress the protocol so as we can be sure that it will operate in the real world Up to now most studies have relied on randomized mobility models 4 30 18 22 especially on the Random Waypoint Mobility Model 31 Thanks to several studies focusing on the wallop of random waypoint mobility model 31 5 27 researchers are now aware of the harmful impact of stochastic mobility patterns One the other hand an effort towards more realistic mobility models can be observed through papers and projects like the Group Mobility Model 14 the Graph based Mobility Model 28 the Obstacle Mobility Model 17 16 the UDEL model 6 7 and the GEMM project 23 Surveys of mo bility models can be found in 9 23 Studying the impact of realistic mobility model on distributed applications Tugcu and Ersoy 29 have shown that the choice of the mobility model has a significant impact on the performance of the mobile systems The effect on the relative performa
26. ation model This section explains how madhoc models the concept of simulation and details some of its important aspects for our point of view in the case of the simulation of MANETs 2 3 SIMULATION MODEL 13 NaturallntegerSensor NaturalNumberSensor NumericalSensor precision int DistributionSensor getUnit org lucci madhoc simulation measure Unit getValueClass class getValueClass class getPrecision int setPrecision int void createMeasureHistory MeasureHistory FA Sensor NumericalMeasureHistory hashCode int getAverage java lang Double MeasureHistory values sequence nonNullValues set getValues sequence createFigure MeasureHistory MeasureHistory org lucci up data Figure getProjection org lucci madhoc simulation projection Projection setProjection org lucci madhoc simulation projection Projection void setSensor Sensor void getSensor Sensor Figure 2 5 The measurement model 2 3 1 Languages All the executable code within madhoc is written in Java Dissimilarly ns 2 and many other simulators use a natively compiled language such as C or C in their core and require the user to write extension in some interpreted language like TCL Nothing like this exists in madhoc Here the only things that are not written in Java are the configuration files if any Then
27. c static fields For each of them it also define some default values Please pay attention at setting all the configuration keys you need at an appropriate value Leaving configuration keys to unappropriate default value is indeed a fairly frequent mistake Configuring a new module If you have developed a new module for madhoc a new protocol a new monitor or whatsoever you might want to give the user the ability to configure it In the case of madhoc as a standalone application the user just had to add new items in the configuration by editing the configuration files In the case of using madhoc as a framework there is no possibility to directly modify the con figuration object nor the its class as the latter is already compiled and contained in a jar file Then the user need to subclass the class of the configuration object and define the new configuration items as fields of the newly created subclass Why using a configuration object In the early versions of madhoc no configuration object existed The user was then required to directly manipulate the classes of the simulator just like developers usually use frameworks like Swing Collections etc But as madhoc evolved it became more and more complex Manipulating the classes of the recent versions of madhoc turns out to be a cumbersome task Introducing the configuration object was then a good way of breaking the complexity of the system by turning the madhoc user s view to a pla
28. ch is out of scope here Then if all the applications that define a termination condition are termi nated the termination condition is true the simulation itself can be said to be terminated As the termination condition of an application depends to itself it must be defined by the programmer 2 3 7 Application deployment In this section we will explain how an application gets deployed within madhoc Deploying the application A monitor is not only in charge of taking measures on the applications its represents It is also in charge of deploying the application accross the network On startup madhoc checks the configuration to get the list of monitors that 2 3 SIMULATION MODEL 17 needs to be created From each monitor class found madhoc instantiates a monitor object and delegate to it the deployment of the application For every node in the network the monitor will instantiate an application object When this is done the application is ready to be executed Taking measures However if nothing more is done the application does not give any feedback to the user of how it works Then the monitor instantiates the sensors that will at each iteration take measures on the application In the case of the graphical model these measures can be displayed at once on the interface In the case of the console mode they will be printed on stdout when the simulation will have finished Keeping an eye In the specific case of the gr
29. d Different mobility models madhoc defines a set of mobility models A mobility model define how nodes move Currently madhoc implements the random mobility model the random waypoint mobility model and the human mobility model The latter is madhoc specific an is highlighted in section 3 2 madhoc s architecture permits node to have distinct mobility models This is achievable by using the simulation in framework mode see section 4 4 2 Figure 2 3 The general mobility framework on top of which all mobility models are implemented 10 CHAPTER 2 MODELS 2 1 2 Software In order to enable node 2 node communication madhoc implements some basic mechanisms which allow the nodes to asynchronously send messages to each others Figure 2 4 The model for the network PHY and MAC layers As explained in section 2 1 1 madhoc models the PHY and MAC layers only in terms of available bandwidth signal power and packet size Collisions in
30. d in section 2 2 2 sensors are asked to take measure on every projections at the end of each iteration Measures are then stored into an history that is taken care of by the projection object The latter history can be retrieve in the following way first gets the instance of the projection we are interested in all the network Projection projection simulation getNetwork getProjectionMap IdentityProjection class and then get the history for the measure we are interested in the throu MeasureHistory history projection getHistory throughputSensor The set of values that have be archived in the history can be obtained by invoking the getValues method on the MeasureHistory object If the mea sure is a numerical measure see section 2 2 2 metrics like the average or the standard deviation can then easily be obtained out of this list 6 4 Accelerating the simulation process If you need to increase the surface of the area you want to simulate or increase the number of nodes in the networks you may want the simulator the operate more efficiently There are several ways of doing this as described hereafter 6 4 ACCELERATING THE SIMULATION PROCESS 49 e Increasing the resolution will fasten up the simulator in a linear way This is the first thing you should look at But increasing the resolution will make the simulation less accurate depending on what you want to simu late Then you should be able to define which is t
31. e the simulation of various types of mobility mobility within mall centers in the city in a train station and so forth In the next section we propose some mobility models built on top of the human mobility model 3 3 Real world mobility based on human mobil ity Mobility impacts the topology on the network but not only The speed of the nodes also have a very important impact on the dynamicity More precisely a network in which the nodes move quickly could perfectly have the same topol ogy than one in which the nodes move slowly meaning that the two networks exhibit the same degree distribution the same number of partitions etc But if the nodes move faster the network is very likely to be extremely dynamic 3 3 REAL WORLD MOBILITY BASED ON HUMAN MOBILITY 25 This section provides the parameters used for modeling various metropolitan networks with madhoc Parameter name Value Spot density Spot radius Wall obstruction ratio human_environment_spot_density human_environment_spot_radius human_environment_wall_obstruction Out spot velocity In spot velocity human _mobility_out_spot_speed human _mobility_in_spot_speed 3 3 1 Highway Parameter name Value Unit Simulation area surface 1000 x 1000 m Mobile phone density 50 node km PDA density 20 node km Laptop density 10 node km Spot density 1 spot km Spot radius 10 10 m Wall obstruction ratio 0 1 Ou
32. eie a a ne ee 3 3 Real world mobility based on human mobility 3 3 A a mel a a D D y ie ares Sieht 913 20 Mall 2 80 2 hela LA a CEREA O OUTECES ia A e AA is 3 3 4 Supermarket Lee 88 x du 8 pee pe eas 3 31 09 Train Station 4 4 4 qa era bbb ee kel ee eed 10 10 10 12 12 13 13 14 15 15 16 16 4 CONTENTS 4 Using madhoc 4 1 Requirements sis ee ee eb eee a hehe 4 2 Packaging io aa ni nenn 4 3 Installation u arras Pp aa 4 3 1 Installation on UNIX operating systems 4 3 2 Installation on Windows operating systems 4 4 Using madhoes 4 40 8 6 does o aa ae a 4 4 1 Configuring madhoc as a standalone application 4 42 Configuring madhoc as a framework 4 5 Using madhoc as a standalone application 5 The graphical user interface 5 1 Monitor views 5 1 1 Standard views 6 Advanced usage 6 1 Creating a new module 6 1 1 Examples sn ea Hi DEE Lee ne 6 2 Graphical and command line user interfaces 6 21 Using the GUL 444 42002 2 aan ee oh et 6 2 2 Graphical 2D plotter 6 3 Using madhoc as a framework 6 4 Accelerating the simulation process 6 4 1 Acceleration tricks specific to broadcasting simulation 7 Open issues 7 0 2 Initialization of the mobility 7 0 3
33. er getIdentifier message setContent id message setSourceStationApplication this getComputer getNetworkingUnit getOutgoingMessageQueue add message for Message msg getIncomingMessages incomingMessageCount 6 1 CREATING A NEW MODULE 45 Obviously the incomingMessageCount field must have been defined Deploying the application The code of the application must now be attached by some means to the simulator itself To do this a monitor for the application must be defined A monitor is a subclass of org lucci madhoc simulation Monitor By deriving this class you just need to give a name to the monitor A name that looks like my_application_monitor is not a bad idea The last thing you need to do is to provide the minimum configuration keys for the newly created application If you want the new application to be the only application to be executed then simply use monitors_class my_package MyMonitor Otherwise add the class name of the monitor to the list of monitor classes This configuration entry makes the simulator aware of the new application The application itself requires a few configuration entries to start Indicate the simulator which application the monitor will deploy my application monitor protocol class my package MyApplication Indicate the termination condition of the application In our case the application never terminates my_application monitor
34. es that are randomly selected This projection is useful if the user wants to operate on only a subset of the simulated nodes for performance reasons for example while keeping a certain degree of statistical confidence 2 3 4 Random number generation madhoc provides a build in customizable random number generator The configuration key random_number_generator_seed allows the user to specify the seed for the random number generator Any integer number is allowed The value ip uses the IP address of the local computer as the seed The value time uses the current time The value time ip uses a mix of both This allows the experiments to be potentially reproducible 2 3 5 Scheduling The application scheduler implemented in the core of the simulator works in a sequential manner This had be chosen for performance issue dedicating one thread by node would have been way too cumbersome and the number of simulated nodes would have been a lot lower 16 CHAPTER 2 MODELS Practically the scheduler manages a list of the simulated nodes It iterates on this list by invoking the run method on every nodes On the use of stochasticity for scheduling Initially the list of nodes on which the scheduler iterated always featured the same order On runtime we could notice that some nodes processed faster that others that some nodes transmitted over the network without any prob lems while other nodes constantly could not We then noticed that on
35. h edge being a street section and each node being a crossroads Similarly to the mall mobility model in order to model the broadness of the streets a variation of the direction of people walking into them is tolerated People walk at a speed randomly chosen between 2 and 6 km h Just like in the random waypoint and mall mobility models people make short pause whose duration is randomly chosen between 1 and 20s The radio propagation model is constrained by the street graph More pre cisely in order to model the signal attenuation of the walls a given node can communicate with other nodes located on the same edge and on aligned edges using the path loss model The broadness of crossroads is represented by allow ing all nodes close to a given crossroads to communicate with one another The pedestrian area mobility model is still under development 8 0 6 Ad hoc computing towards the ad hoc grid Our main target is the investigation of computations distributed over mobile ad hoc networks madhoc were developed with this idea in mind We are currently working at realistically modeling the properties of the fu ture ad hoc grid and at predicting its behavior We are now focusing on building a basic grid application which consist of the distribution of on single task It may sound simple but actually many questions arise According to what policy the distribution should occur How to ensure that the result of a task will come back the client node
36. harm can be simply ignored e Making the nodes moving faster is also a broadcasting specific trick It increases the number of connection disconnection which permits some broadcasting protocols like DFCN and AHBP EX which use the new connection event as a trigger for considering a re emission to process significantly faster 50 CHAPTER 6 ADVANCED USAGE Chapter 7 Open issues madhoc targets the simulation of metropolitan networks So far only few re search projects have focused on this Along the conception development of madhoc we have been facing several complex issues 7 0 2 Initialization of the mobility The initialization of mobility is difficult How can we make sure that the initial network is valid The structure of the network is a consequence of the mobility But because the mobility rules are complex the topology it will generate cannot be predicted A solution is to run the simulator by discarding the first measures How long this initialization process should last This issue is a real problem because it might prevent researchers from build ing relevant models for certain mobility patterns 7 0 3 Altering the resolution of the simulation The concept of resolution is described in section 2 3 2 madhoc has initially been developed considering a fixed resolution of 4 iterations per second Lately we had to change this by supporting a user defined resolution But simulating a network with a resolution that is no
37. he maximum resolution under which the behavior of the simulation will change insignificantly However you should be extremely cautious by altering the resolution of the simulation because it may lead to unpredictable results e Reducing the number of projections also improve the performance of the simulation It is important that you define only the projections you need for your simulations Adding discarding projections has no effect on the simulation engine e reducing the sensors by selecting those only which provide the metrics you are interested in Several sensors like the ones which counts the number of partition which measure the degree distribution in the network etc require a considerable amount of time to proceed If you are not interested in their measures just discard them 6 4 1 Acceleration tricks specific to broadcasting simula tion e Discarding the RAD is a trick specific to broadcasting If the resolution of the simulation is high the RAD becomes useless because it is a low level mechanism In some broadcasting protocols such as DFCN AHBP etc the RAD is important because it has a serious impact on the order protocol level information is delivered to the node Hence discarding the RAD may change the behavior of the broadcasting protocol Discarding the RAD in DFCN makes it loose some benefit of its cumulative neigh borhood mechanism which head to more message emissions Depending on what you want to simulate this
38. he rendering of numeric data It may be used for scientific statistic geometric data rendering More generally it is able to render any object that can be said to be composed of points Up keeps things easy using a very common object oriented model the com posite design pattern Practically Up merely renders a figure assuming that a figure is composed of points and child figures Up also automatically adapts the axis system according to the graphical environment EPSGraphics is a package that allow Java application Up in our case to pro duce PostScript files It is required only if you want to generate PostScript files iText is a package that allow Java application Up in our case to produce PDF files It is required only if you want to generate PDF files Javamail is not mandatory but it allows madhoc to send me a e mail message in case it encounters an problem This is totally transparent to the user Javamail is Sun product and is freely downloadable from Sun s website Note that Javamail requires the Sun s activaction toolkit which is also free 4 2 Packaging madhoc comes as a set of jar files The jar files which come with the madhoc distribution contain both the source code and the compiled bytecode of mad hoc In the distribution you have the source code might be obsfucated for 31 32 CHAPTER 4 USING MADHOC performance reasons If it is not the case feel free to unjar the source and look at it The following list g
39. ity 50 000 node km PDA density 30 000 node km Laptop density 500 node km Out spot velocity 0 3 1 5 m s In spot velocity 0 1 0 5 m s 28 Probabilty in 0 1 CHAPTER 3 SCENARIOS Supermarket Figure 3 8 On the left an example of a network formed by people shopping in a supermarket On the right the average degree distribution for such networks 3 3 5 Train station The train station environment model pedestrian walking in the hall of a train station going to shops buying tickets The resulting movement for the node is slow with lots of pauses Parameter name Value Unit Simulation area surface 100 x 100 m Spot density 10 000 place km Spot radius 1 2 m Wall obstruction ratio 0 1 Mobile phone density 8 000 node km PDA density 2 000 node km Laptop density 500 node km Out spot velocity 0 3 1 m s In spot velocity 0 3 0 3 m s 3 3 REAL WORLD MOBILITY BASED ON HUMAN MOBILITY 29 Probability in 0 1 Figure 3 9 On the left an example of a network formed by people roaming in a train station On the right the average degree distribution for such networks 30 CHAPTER 3 SCENARIOS Chapter 4 Using madhoc 4 1 Requirements madhoc requires the version 1 5 of the java runtime environment This software is freely available on Sun s website It also makes use of the following projects Up isa Swing component dedicated to t
40. ive some explanations on the content and role of each jar file Madhoc jar is the core of the madhoc simulator which is absolutely necessary for running it Madhoc Broadcasting jar contains the implementation of standard broad casting protocols Madhoc HumanMobility jar contains the code for the mall mobility model 7 Madhoc NetworkMonitor jar contains the code sensors graphical views etc for monitoring the general behavior of the network Madhoc AdHocComputing jar contains the code for ad hoc scattering which is the reverse operation to broadcasting Madhoc BroadcastingMalaga jar features some specific code developed for the need of the researchers at M laga university Toools jar is a set of utility classes used by madhoc Up jar permits the 2D graphical representation of the data collected by the sensor plus various data structures Up is a standalone project hosted at http amy sunsite dk up skinlf jar permits the use of the Skin Look amp Feel which is required by some version of madhoc This jar file is needed only if you use the graphical interface and want to have the use this look zfeel 4 3 Installation As madhoc is a pure Java application installing it on any computer is trivial madhoc comes with a set of jar files These files can be freely downloaded at http www lih univ lehavre fr hogie madhoc Please take care at downloading the most recent files which can be identifiable by looking a
41. jections feature the same tools called views These views are organized are classified using the tree on the left side of the window with regards to the monitor which made them available One or several views can be activate at the same time by clicking on them for activating multiple views click on them by pressing the Ctrl key Numerical measure list This view lists all the sensors available in the simulator The values presented are the ones extracted by considering the given projection only So the values for the same sensors on different projections may be different The measures are classified according to the monitor that made them available 6 2 2 Graphical 2D plotter This view allows the graphical representation of any measure in relation with any other In order to define which measure to represent as a 2D plot select them using the numerical measure list view Both views must hence be active in order to do that 6 3 Using madhoc as a framework By using the framework the user actually the programmer needs to define the configuration by setting the values of the public fields of a ConfigurationKeys object ConfigurationKeys keys ConfigurationKeys show how to set a parameters 6 3 USING MADHOC AS A FRAMEWORK 47 there are plently of such parameters available to the user keys simulation_name my simulation As explained in section the configuration is a set of key values pair At the level of the
42. ly their im importance in military applications is tightening but also their impact on business is becoming clear The wide spread of mobile phones PDAs Pocket PCs which now embed Bluetooth and WiFi IEEE 802 11 network adapters enables the spontaneous creation of metropolitan ad hoc networks 10 These networks could then constitute the infrastructure of numerous appli cations such as emergency and health care systems 19 groupware 8 gaming 25 13 24 advertisements customer to customer applications like the UbiBay project 12 etc Up to now MANETs have been mostly investigated at the lowest layers of the network stack Indeed many studies tackle issues like routing 26 broad casting 30 security etc The vast majority of them resort to MANETs simulation Unfortunately the simulation platforms available do not model the networks in a realistic manner and are limited For example by simula tions broadcasting 30 has shown to operate well on non partitioned networks composed of few identical stations moving according the random waypoint mo 5 6 CHAPTER 1 INTRODUCTION bility model But how would broadcasting protocols operate on large gt 10 000 stations heterogeneous networks More generally is broadcasting merely prac ticable over this structure Such questions are still unanswered because no tool actually enable researchers to solve them Unfortunately the investigation of metropolitan ad hoc network
43. more than providing convenient facilities for the applications to obtain in formation about their changing environment Connection disconnections In order to enable the nodes to be aware of new neighbors simulators usually use event driven notification More precisely when the simulator finds a new link it dynamically invokes a method on the two nodes involved This method initially abstract needs to be implemented by the application that defines what to do in case of a connection disconnection event This application of the ob server design pattern based on polymorphism has proved elegant and flexible However madhoc does not uses this because based on our experience of ad hoc programming we assumed that event based strategies bring more complexity than flexibility madhoc then does not dynamically invoke anything on nodes newly in range Instead it provides all nodes with an utility method that re turns a set of the nodes that were not yet in range last time it was invoked Applications are then free to invoke this method or not Note Another drawback of the event based notification is that the code of handler methods is executed by the simulator in a routine dedicated to the construction of the interconnection network a routine that should not execute any application level code Application identification As nodes can execute several applications there is the need to have a way of identifying the different applications running
44. mply selecting the requested one on the tree widget note that selecting a tree node at zero and first level does not do anything You can select multiple views by keeping the Ctrl key pressed while clicking on the corresponding leafs this way of doing is very common on graphical user interfaces 5 1 1 Standard views madhoc comes with a set of standard monitors that features views which are directly usable Before starting madhoc ensure that the configuration key monitors_class include the org lucci madhoc broadcast NetworkMonitor and org lucci madhoc simulation monitor SimulationMonitor classes Aircraft views The aircraft views comes with the network monitor It provides an 2D repre sentation of the simulated network as if it is was observed from an aircraft or a satellite Image 5 3 shows and example of an aircraft view 5 1 MONITOR VIEWS 39 Figure 5 3 An aircraft view of a network Numerical measures list The numerical measures list view comes with the simulation monitor It provides a table whose rows represent a numerical measure see section measures All numerical measures in the simulator are represented here As shown on image 5 4 different colors permit to recognize which monitor the measure come from the colors used here are the same as the colors used on the tree widget on the left side of the projection window buna 7 Pe v n 25211575F u BAA15763 J 18 150863727 a
45. nar one 4 5 Using madhoc as a standalone application The executable class that must be invoked for starting madhoc is org lucci madhoc Madhoc 36 CHAPTER 4 USING MADHOC There are two ways of invoking it You can specifiy a directory and or configuration files in the command line Configuration files will be loaded directly while directories will be scanned for configuration files to be loaded java org lucci madhoc Madhoc config a madhoc b madhoc Then the following command loads all the configuration files in the current directory java org lucci madhoc Madhoc If you do not specify anything on the command line madhoc reads its con figuration for its standard input You can then invoke madhoc this way cat config madhoc java org lucci madhoc Madhoc Warning In this case you should however make sure that all the configura tion files start or end with a newline character If not by concatening two files whose the first does not end with a newline character the pipe will append the first line of the second file to the last line of the first file and will hence generate an invalid syntax Chapter 5 The graphical user interface If you wish to launch the graphical user interface you need to set the simulation_interaction_mode to graphical This will completely initialize the simulation and then instanti ate a graphical application that will give the user the ability to observe in many ways the simulation process
46. nce becomes more important especially when the algorithms try to predict the mobility of the nodes madhoc currently defines two mobility models random waypoint and human mobility The following table gives the name of the configuration keys that need to be set in order to define a given value for the simulation Parameter Configuration key Simulation area surface simulation_area_surface Mobile phone density network_phone_density PDA density network_pager_density Laptop density network_laptop_density Hotspot density network_hotspot_density 3 1 Random waypoint Random waypoint is a commonly used mobility model In the random waypoint mobility model as described in each station chooses randomly a destination 19 20 CHAPTER 3 SCENARIOS in the simulation area and moves towards this destination with a randomly chosen velocity When the destination is reached the station remains at the same place for a while This process is repeated by each station until the end of the simulation In addition to that within madhoc in order to reduce the harmful behavior inherent to the random waypoint mobility model 31 each simulation process is initialized using the stationary distribution pattern 21 Random waypoint is supported in madhoc for the sake of compatibility and respect of the previous works in the domain Parameter name Configuration key Node velocity interval Pause time inte
47. network formed by people wandering accross in a market On the right the average degree distribution for such networks Upon time their target change Most of the time people have a dynamically changing list of targets This list can be composed of two spots like a secretary regularly moving from her office to her boss office and vice versa or more people shopping This list of target spots is dynamically changing because of various parameters that will appear upon time locations of the target places closure times high frequentation times etc The behavior described in this paragraph is different from what is proposed by random waypoint Our scenario can be said to be some constrained waypoint the constrained waypoint mobility model does not rely on randomness then the destinations and pause times are given by a scenario enriched with advanced rules based on human s decision The human mobility model proposed by madhoc models this The general principles of human mobility The human mobility defines that the simulation area is populated with fixed places that constitutes targets for the nodes in move Places are randomly located within the mall so as the distance between two given places must not be lower than 10m A place has a round shape whose radius is randomly chosen in a given interval surrounded with walls Walls are important because they obstruct mobility and attenuate the radio signal People within a given place move
48. o pay attention to the fact that Cygwin does not runs a UNIX JDK but a Windows one Consequently the PATH separator the user need to specify on in Cygwin configuration is The way Cygwin manages the file system poses also some troubles For example it turns out to be better to write c then cygdrive c which theoretically both refer to the C logical disk drive AA Using madhoc madhoc can be used in two different ways as a framework or as a standalone application On the one hand using it as a framework consits of manipulat ing its API application programming interface from a user program Doing this requires a advanced knowledge of the Java programming language and a good understanding of madhoc object oriented model On the other hand using madhoc as a standalone application consists of executing the madhoc applica tion just like any other application within a process of the operating system In this case madhoc works pretty much like a Unix filter it accepts its input data from the command line or as a command line argument and generate its output on the standard output or to a file Whether madhoc is used as a framework or as a standalone application it needs to be fed with some configuration information This information defines what to do and how to do it 34 CHAPTER 4 USING MADHOC Formally speaking the configuration is a set of key value entries in which each key is a unique string and the value is a set of strings
49. on one given node madhoc defines than applications running on a given node are identified by their class similarly IP identifies processes by their port numbers Consequently the situation in which two applications of the same class would run on the same node conflicting services cannot happen Node 2 node communication Application communicate exclusively in asynchronous and unconnected mode by using the message passing paradigm madhoc does not implement any connected mode connection patterns such as sockets A message can be sent in unicast one recipient multicast multiple recipients and broadcast all nodes in the neighborhood are implicitely recipients modes madhoc does not strongly define what a message is It only impose that a message should be transferable across the network links Then by defining a new kind of mes sage the user should implement some specific method that will allow madhoc 12 CHAPTER 2 MODELS to know the number of bytes used to code the message and to make a copy of the message 2 2 2 Monitor Ad hoc applications are inherently distributed over a highly dynamic network infrastructure Consequently deployment and monitoring are challenging issues Simulation allows to put in place things that would not be possible in the real world Then madhoc defines the self explanatory concept of monitor Definition A monitor is an abstraction defined at the level of the simulation It does not have any
50. or such networks 3 1 3 Concert place Parameter name Value Unit Simulation area surface 50 x 50 m Mobile phone density 5000 node km PDA density 1000 node km Laptop density 0 node km Velocity interval 0 0 2 m s Pause duration interval 5 600 s The concert place and the market place environment lead to a similar con nectivity This is because of similar densities The main difference between these two environment resides in the velocity More precisely while nodes in a market place have the speed of a pedestrian those in a concert place barely move This difference of velocity of the nodes is not visible on static views as these presented here but has a great impact on how protocol process In the specific case of a dynamic broadcasting protocol the movements in the market place environment which result to many connection disconnections between nodes may entail lots of re broadcasting of a packet 3 2 Human mobility Human mobility is the default mobility model in madhoc It is a generic mobility model that roughly represent the clever mobility of people in metropolitan areas 22 CHAPTER 3 SCENARIOS e we e e 4 i w 2 s e ee eg i e a di ee dl gt D E a e o N et gt oo s gt k ar s R vet e e q ee E 7 hi gt x A s de i e y oo e s e s mn
51. outing protocols for ad hoc mobile wireless networks In IEEE Personal Communications 1999 27 Christian Schindelhauer Tamas Lukovszki Stefan Ruhrup and Klaus Vol bert Worst case mobility in ad hoc networks In SPAA 03 Proceedings of the fifteenth annual ACM symposium on Parallel algorithms and archi tectures pages 230 239 ACM Press 2003 28 Jing Tian Jorg H hner Christian Becker Illya Stepanov and Kurt Rother mel Graph based mobility model for mobile ad hoc network simulation In Annual Simulation Symposium pages 337 344 2002 29 Tuna Tugcu and Cem Ersoy How a new realistic mobility model can affect the relative performance of a mobile networking scheme Wireless Communications and Mobile Computing 4 4 383 394 2004 30 B Williams and T Camp Comparison of broadcasting techniques for mo bile ad hoc networks In MOBIHOC Proceedings of the ACM International Symposium on Mobile Ad Hoc Networking and Computing pages 194 205 2002 31 Jungkeun Yoon Mingyan Liu and B Noble Random waypoint considered harmful In INFOCOM Twenty Second Annual Joint Conference of the IEEE Computer and Communications Societies pages 1312 1321 March 2003
52. processors will run n simulation servers There are no relations between simulation servers running on a same given computer they operate in an autonomous and independant manner On every computer participating to madhoc s grid there is a server running This server periodically looks for tasks to execute The number of task it can execute simulatenously depends on the number of processor of its hosting com puter As a madhoc simulation is a monothread processus a single computer can simulatenously run as many simulation as it embeds processors 57 58 CHAPTER 10 GRID Bibliography 1 Y 10 GNUPlot http www gnuplot info Ultimate Plotter Up http amy sunsite dk up Kaizar Amin Gregor von Laszewski and Armin R Mikler Toward an Ar chitecture for Ad Hoc Grids In 12th International Conference on Advanced Computing and Communications ADCOM 2004 Ahmedabad Gujarat India 15 18 December 2004 Christian Bettstetter Smooth is better than sharp a random mobility model for simulation of wireless networks In MSWIM 01 Proceedings of the 4th ACM international workshop on Modeling analysis and simulation of wireless and mobile systems pages 19 27 ACM Press 2001 Christian Bettstetter Giovanni Resta and Paolo Santi The node distribu tion of the random waypoint mobility model for wireless ad hoc networks IEEE Transactions on Mobile Computing 2 3 257 269 2003 Stephan Bohacek and Vinay S
53. ridhara The graph properties of manets in urban environments In In Submission 2004 Stephan Bohacek and Vinay Sridhara The udel models manet mobility and path loss in an urban In In Submission 2004 Dominik Buszko Wei Hsing Dan Lee and Abdelsalam Sumi Helal De centralized ad hoc groupware api and framework for mobile collaboration In GROUP 01 Proceedings of the International ACM SIGGROUP Con ference on Supporting Group Work pages 5 14 ACM Press 2001 T Camp J Boleng and V Davies A Survey of Mobility Models for Ad Hoc Network Research Wireless Communications amp Mobile Computing WCMC Special issue on Mobile Ad Hoc Networking Research Trends and Applications 2 5 483 502 2002 Marco Conti Silvia Giordano Gaia Maselli and Giovanni Turi Mobile man Mobile metropolitan ad hoc networks In Proceedings of the Sth International IFIP TC6 Conference Lecture Notes in Computer Science LNCS 2775 pages 194 205 September 2003 59 60 BIBLIOGRAPHY 11 Hannes Frey Daniel G rgen Johannes K Lehnert and Peter Sturm A java based uniform workbench for simulating and executing distributed mo bile applications Scientific Engineering of Distributed Java Applications november 2003 12 Hannes Frey Johannes K Lehnert and Peter Sturm Ubibay An auc tion system for mobile multihop ad hoc networks In Workshop on Ad hoc Communications and Collaboration in Ubiquitous Computing Environ ments 2002
54. rval random_waypoint_mobility_velocity_interval random_waypoint_mobility_pause_interval 3 1 1 City place Parameter name Value Unit Simulation area surface 200 x 200 m Mobile phone density 3 000 node km PDA density 500 node km Laptop density 100 node km Velocity interval 0 3 1 m s Pause duration interval 0 10 S As illustrated on figure 3 1 the network which is formed by the presence of people walking accross a city place Because a city place is an open area no obstacle prevent the propagation of radio waves Consequently the network is well connected The devices not equipped with a WiFi adapter tend to be isolated due to their short coverage radius 3 1 2 Market place Parameter name Value Unit Simulation area surface 200 x 200 m Mobile phone density 5 000 node km PDA density 1000 node km Laptop density 10 node km Velocity interval 0 3 1 m s Pause duration interval 0 01 60 s Figure 3 2 shows the result in terms of node location of the market place environment As illustrated in figure 3 3 the resulting network is heavily connected This comes because of the high density of devices and the lack of obstacles 3 2 HUMAN MOBILITY 21 City place Probability in 0 1 Figure 3 1 On the left an example of a network formed by people wandering accross a place in a city On the right the average degree distribution f
55. s has an even greater need of simulating Indeed investigating ad hoc networks poses a fundamental problem due to the lack of existing networks carrying out experiments on real devices is not pos sible At best this could be done on a small set of devices but doing this would not help much in producing relevant results The only workable alternative turns out to be software based MANETs simulation Unfortunately simulation does not really solve the issue most simulators do not feature realistic network models in terms of station mobility and propagation of the radio waves and do not permit the simulation of sensibly large networks For example the most popular mobility model random waypoint has been found by Yoon and al 31 to have harmful side effects on the simulation Efforts towards realistic models have been observed only recently through projects like the NAB 15 and Jane 11 20 simulators the UDEL 7 models etc These initiatives all attempt to allow the simulation of metropolitan ad hoc networks Unfortunately these are on going research projects they are hence hardly re usable as is Therefore the studies of metropolitan MANETs may require the development of custom simulators Our initial purpose is the investigation of the ad hoc grid issue By looking for an experimentation platform we realized that developing a custom solution was the most sensible choice hence this article This article presents the madhoc simulator m
56. t its date which is part of its filename Copy these files anywhere in your file system and include them in the class path At this step madhoc is completely installed on your computer 4 3 1 Installation on UNIX operating systems On UNIX type operating systems including Linux the easiest way of installing madhoc as well of any pure java application is to copy the jar files in your HOME 1ib directory Then the HOME profile file or another file depending on your shell should have the following command which automatically include 4 4 USING MADHOC 33 into the CLASSPATH environment variable all the jar files found in the HOME 1ib directory for j in HOME lib jar do CLASSPATH CLASSPATH f done export CLASSPATH 4 3 2 Installation on Windows operating systems There is no de facto standard for installing Java programs on the Windows operating system The installation of madhoc will then depend on your java runtime environment Note that on Windows the PATH separator character which seperate the jar file paths within the CLASSPATH variable is the character on UNIX the path separator is Installation on Cygwin environments Cygwin is an UNIX environment built on top of the Windows platform It is a very good option to those who do not wish to set up a dual boot system and turns out to me more flexible solution than the latter Installing madhoc on Cygwin is no problem The user simply got t
57. t spot velocity 20 40 m s In spot velocity 20 40 m s The highway mobility environment is characterized by nodes moving at a very high speed The density of spot is defined so as there are preferably only 3 spots in the simulation area Numerous spots would result in too many in tersections of nodes which would lead to unrealistic behavior environment is very different from the other ones in the sense that the con nectivity it produce it specific most links are organized into chains of nodes moving in opposite senses 3 3 2 Mall The highway The mall mobility model represents the mobility of nodes with an area made of shops connected by corridors Spot model shops Parameter name Value Unit Simulation area surface 300 x 300 m Spot density 1000 place km Spot radius 1 10 m Wall obstruction ratio 0 7 Mobile phone density 5000 node km PDA density 1000 node km Laptop density 500 node km Hotspot density 0 node km Out spot velocity 0 3 1 m s In spot velocity 0 3 0 8 m s 26 Probabilty in 0 1 be CHAPTER 3 SCENARIOS Highway 10 Degree number of nodes 12 Figure 3 5 On the left an example of a network formed by cars driving on a highway section On the right the average degree distribution for such networks gt 4 x IRA y z4 W a T il Va T amp X p y gt B 2 e lt 3 Pr d 4 F y y y L Lay Y ESL RP
58. t static is extremely difficult The most difficult point is probably the simulation of the data transfer which is tightly linked to time 51 52 CHAPTER 7 OPEN ISSUES Chapter 8 Targetted applications 8 0 4 Broadcasting madhoc were primarily designed to make easy the development and experimen tation of broadcasting protocols Two efficient protocols were defined using madhoc DFCN and CABP DFCN is a bandwidth efficient protocols for mo bile networks CAPB is an extension of the DFCN protocol CAPB adapts its strategy to the degree of urgency of the message that is being broadcasted The more urgent is a message the less greedy will CABP react 8 0 5 Mobility models Mobility models A second application of the madhoc simulator is the development of novel mo bility and radio propagation models for the simulation of metropolitan networks Then we have developped the mall mobility model The mall mobility model simulates the mobility of people walking in a mall center It defines a mall as a wide area made of shops and corridors which connect them People in the mall then move from shop to shop using the corridors The mall mobility model uses the constraint waypoint mobility model at two different scales at the levels of the shops and at the bigger scale the level of the mall First a shop is a round place whose radius is randomly chosen in a given interval surrounded with walls which obstruct human mobility and attenuate
59. terferences are not modeled by statistical models This makes the simulator lightweight and consequently enables the simulation of large networks In the code madhoc does not make any clear difference between PHY and MAC layers Their properties are both expressed within the same algorithms Network and transport layers madhoc consider that in the context of large heterogeneous MANETs multi hop networking is unachievable Then it does not feature any internal mechanism for multi hop applications like routing This limitation is not a problem as madhoc aims at simulating localized applications 2 2 Ad hoc application model 2 2 1 Application Definition There is not yet any universally accepted definition of what is an ad hoc ap plication In our context an ad hoc application is a piece of software that is executed in the mobile devices and is able to communicate in its steadily chang ing environment The application is not the global thing that run on all nodes 2 2 AD HOC APPLICATION MODEL 11 in a network it is the piece of code that run on one particular node The appli cation can be executed on only a subset of the nodes in the network Moreover several applications can be executed on one single node Because the environment is changing the application must then be robust to connections disconnections of peers As this robustness depends on the ap plication it cannot be defined at a middleware level The middleware cannot do
60. termination conditions none Furthermore the application does not need specific code to be initialized my application monitor initializer none We did not define any measure for the application so filtering measure makes no sense Just allow everything or nothing it does not matter my_application monitor measures regex The application wants to know how often it must send hello message This is given by the following configuration entry In this case a message is sent every 2 seconds hello_application_frequency 2 46 CHAPTER 6 ADVANCED USAGE 6 2 Graphical and command line user interfaces madhoc can be used in 3 differents ways as a console program as a graphical application as well as a framework In either case the user needs to define the configuration for the simulation Whether you want to use the graphical user interface or the command line tool you need to provide the configuration for the network you want to simulate Their entry point is the same that is the org lucci madhoc Madhoc class Whether madhoc route the execution to the GUI or the console mode depends on the simulation interaction mode configuration key You can set it either to graphical or console 6 2 1 Using the GUI The graphical user interface aims at providing the user with excellent visualiza tion capabilities on the application he is developing Organized as a desktop each internal window represent a projection of the network All pro
61. used on each projection generating several measure repository Several projections has been predefined As convenient as projections are processing the projections is a cumbersome task for the simulator at each iteration it needs to find out which nodes belong to a projection and then apply the measurement available on this newly created set of nodes It is then important that the user define only the projections that are relevant to him madhoc has a set of pre defined projections Some of them are described in the following 2 3 SIMULATION MODEL 15 Figure 2 6 The model of network projections identity projection gathers all the stations in the initial network This is the most basic projection as it projects everything window projection considers a sub area of the simulation area whose center is the center of the latter The window can be either a square or a circle The size of the window is parametrizable network technology based projection considers the subset of the nodes in the network that embeds a network adapter of a given type only connected nodes projection gathers all the node that have at least one neighbor random projection is a projection made of a set of nod
62. want to use the simulated time measure on the X axis image 5 6 p Swan XY Figure 5 5 The widget that select the dimension to be set Figure 5 6 The evolution of network coverage in the case of a broadcasting application upon time Distribution view The distribution view an example is given on image permits the 2D reprenta tion of every distribution measures 2 2 2 available in the current simulation The export function described in section 5 1 1 is also available here Image exportation Often it turns out to be very useful to export the displayed image to a disk file For doing this right click on the rendered image A popup menu will 5 1 MONITOR VIEWS 41 then propose a Save as menu item that will give you the ability to save the image to various common image formats If the image is targeted to an scientific paper vector format PostScript PDF or SVG is obviously the way to go as vector based image representation ensure perfect scaling or the produced image People willing to re work the rendering with a tool like GNUPlot 1 have the possibility to export the rendered data as a tab separated file that will be easily red by GNUPlot If the image is targeted to a web site bitmap format will be preferred Note that JPG is a lossy compression technique that is better forget about in our concern since it produce dirty graphics This functionality as well as the rendering capability is provided

The Madhoc Metropolitan Adhoc Network Simulator

Contents

Download Pdf Manuals

Related Search

Related Contents