Home

Quick deployment network using MANET

Contents

1. Node D receives OGM from node A Now it knows the existence of A Node A sends A m his OGM to the 5 neighbours Figure 5 1 OGM propagation example 5 1 INSIDE BMX6 25 5 1 2 Protocol data structure Unlike most of other mesh routing protocols BMX6 is not using IP addresses as node identifiers Instead global identifiers SHA2 based Description Hashes DHASHs and very short local identifiers are used to represent nodes and related interfaces in the mesh As a result of this approach the IP address size only has a marginal impact on the total protocol data overhead Figure 5 2 shows the low overhead impact of this implementation p 200000 T TT 05 505 1005 1505 2005 2505 3005 3505 4005 Figure 5 2 Comparative between BMX6 and OLSRd overhead in a virtual environment with 39 nodes Y bytes X seconds Red OLSRd 0 6 1 Blue BMX6 BMX6 organizes packets in the following structure e Each Packet is wrapped into an IP UDP header e The UDP header is followed by BMX6 packet header of 17 bytes e The remaining part of a packet is filled with one or multiple frame Content Link Header IPv6 UDP BMX6 routing protocol packet Size Bytes 40 8 A frame consists of a frame header and frame data The frame header defines the size and type of the following frame data A list of currently existing frame types is given in the next table The size of the frame h
2. option key tmp qmp_key e community node id An number identifying the node In this case node name will be qmp901 If not specified it will be auto configured based on the last byte of the MAC address e key Path to file containing the QMP key It is a hash generated randomly on each boot used by some QMP applications for security purposes 7 3 2 Interfaces Interfaces section is very simple as it consists in three options which indicate the operation mode for each physical network device as shown in the next example config qmp interfaces option lan devices ethO wlani option wan device eth1 option mesh devices wlanO wlan2 e lan devices Interfaces used for final client connection a bridge between them will be created and the DHCP server will be enabled e wan device Interface connection with another network A DHCP client will be enabled in this interface to automatically get the network configuration e mesh devices Interfaces used for meshing Several VLAN will be created and routing protocols will be enabled 7 3 3 Networks In this section all options related with networking are found Let s see an example config qmp networks option dns option lan address SIT 2330 2271 option lan netmask 2255 255 255 0 option niit_prefix96 ia 020 option mesh protocol vids 2olsr6 1 bmx6 2 option mesh vid offset
3. 10 40 CHAPTER 7 INSIDE THE QMP DEVELOPMENT option bmx6 mesh prefix48 02 0 0 option olsr6 mesh prefix48 fd01 0 0 option babel mesh prefix48 fd03 0 0 option netserver 1 e dns Nameservers used for the device and final clients e lan address IPv4 address used for final client bridge interface see 7 3 2 e lan netmask IPv4 netmask used for final client bridge interface e niit prefix906 Niit is used for 4in6 tunneling see 7 2 4 e mesh vid offset VLAN tag offset used for routing protocol interfaces e eth0 10 e mesh protocol vids VLAN tag suffix added for routing protocol interfaces mesh vid offset mesh protocol vids e xxxx mesh prefix48 First 48 bits used for IPv6 calculation see IP STRUCTURE CHAPTER e netserver Enable disable netserver used for bandwidth tests 7 3 4 Wireless To configure WiFi devices QMP offers an abstraction layer which intends to be more easy and intuitive than the one offered by OpenWRT The user only needs to fill some fields and QMP will do the rest Note that configuration is different for layer 3 networking saw in section 7 3 3 and layer 2 this one Ad Hoc mode is not strictly related with Mesh routing user can choose AP Client as link mode and probably WDS in the future to run routing protocols too if needed Wireless section consists in two different set of options General section affects all WiFi devices looks like that confi
4. 3 SMS JSON plug in 26 6 OpenWRT a basis for QMP 27 6 1 Brief introduction to OpenWRT 27 6 2 Why Open WRI ee Ro Rem ee nn d 28 6 3 QMP as part of OpenWRT 28 64 OMP main packages 44 245 8 2a 2 a upa wa O ve Y 4s 29 6 4 1 qmp smal node u 2 2 2222 mn 30 6 42 qmpesbie node au 2 24 Al aa am ie ee oe e 30 7 Inside the qMp development 31 Al The TP structures 0 RR pa Be re cR eer a e 33 TAL ULA address S s ox a eR 20 er An Se Ee ase dus 34 T4 2 TANA addresses 2 eumd T eg 35 7 1 3 IPv4 addresses Q u m mox y eon ROSA LR we ek SR UR ees 35 7 2 The interfaces structure 36 1 2 L Mesh interfaces sx eu g dk atin ev ed Bodo e e ed 36 7 2 2 AN interfaces i o 6 06 2404 ke er een 36 7 2 3 WAN interface eade p arag iiaa e a A a E pa A a an bn aOR e ao 36 7 2 4 NIIT 4to6 6to4 interfaces 37 7 2 5 BMX6 tunnel interfaces 37 7 2 6 An example of 1 37 7 3 UCI configuration 39 7 3 1 39 id Interfaces Sed bo er oko Boe S k DOO ee I qo ond e
5. Build hardware and install software 5 2 Perform stress tests 5 3 Fix detected problems 6 Tests in real environment Dir 25 Dev 75 6 1 Fix detected problems Documentation phase 1 Project documentation Dir 7596 Doc 25 2 User manual Dir 10 Doc 90 3 Administrator manual Dir 1096 Doc 90 Broadcasting phase 1 Lecture for interested people Dir 5096 Res 50 2 Mailing lists for users and developers Dir 10 Doc 9096 3 Web site Dir 5096 Dev 50 Budget The budget needed to perform the project Note that Director and Responsible are volunteers so they are not getting any money from this budget Reason Description Amount Director 102h 0 h 0 Responsible 46h 0 h 0 Developer 413h 15 h 6195 Documentor 254h 12 h 3048 Hardware 8 nodes 423 node 3384 Association Taxes 18 VAT 2306 34 Total 14933 34 3 4 MY CONTRIBUTION 13 3 4 My contribution In the last three years I have been working as volunteer for the Guifi net project Wireless Community mainly in my current district Sants from Barcelona I started with some other people a new project called GuifiSants net to bring the Guifi net coverage to the district non existent before To do it we decided to use Mesh MANET structure using the firmware developed by the GSF folks That s how I started to work in this kind of networks I have
6. also be given to the final clients personal computers they cannot be used to access Internet but to surf over the Mesh network An IPv6 address has 128 bits we have chosen the next distribution to calculate the digits MAC or clients 64 X Interface 12 4 Content ULA prefix Bits 8 QMP prefix Cloud 32 8 This is the meaning of each IP block Content Description ULA prefix The standard ULA prefix fc QMP prefix We are currently using three prefixes to define the network as QMP network and to differentiate between the different protocols They are statically configured always the same Cloud The number of QMP cloud If several clouds are physically separated but exist the possibility of joining together the number should be different for each one Using 8 bits we are able to define 256 clouds For interfaces used to meshing X 000 where routing daemon is running For the interface where the clients are connected X is randomly calculated when needed and checked with the whole network to avoid IP collisions Interface The index of the interface used inside the node For instance eth0 will be 0 and wlan0 1 We are able to have 16 interfaces in the same device For the clients interface these bits are random so in fact they are part of the X MAC The last 64 bits will be calculated according the MAC address of the device following EUI 64
7. be deprecated Batman adv a layer 2 protocol This is probably the most different protocol because it is operating on MAC layer 2 It can be imagined like a big smart virtual switch where all nodes are connected together Nowadays it is included on kernel as a module so to use it we just need to load the module and use brctl userspace tool to control it Currently the development is managed by Marek Linder one of the BATMAN originators 18 CHAPTER 4 SMALL RESEARCH IN BATTLE OF MESH BMX6 BATMAN experimental version 6 It is based on the code of BATMAN 0 3 but now in the version 6 it is totally rewrote Initially it was a branch of BATMAN maintained by Axel Neumann with some improvements in the routing algorithm It is probably the most close protocol to the original BATMAN using UDP flood to discover neighbours and avoiding link state idea Some extra features have been developed like a short text message plugin which permits to send any information to the rest of nodes using the same routing packages 4 3 1 Technical details or our testbed nodes have installed and configured a total of 5 routing protocols each of them running in different networking range Batman Babel OLSR Batman adv BMX6 10 10 101 0 24 10 10 102 0 24 10 10 103 0 24 10 10 104 0 24 10 10 105 0 24 We classify the protocols in two diferent groups e A Babel OLSR BMX6 e B Batman Batman adv That is becau
8. been involved in the project QMP from the beginning November 2010 but not in project drafting I was one of the three developers and I contributed in some documentation tasks Specifically I have been focused mainly in the next points e auto configuration system e web management interface e shell management tools e software development kit e hardware nodes building e technical documentation But the most important work from my point of view has been the decision making The organization of work during the project execution has been totally horizontal so all involved people have been able to participate in all decisions taken We have had several meetings and mail exchanges discussing about all points related with the project Currently the funded project has been finished we deliver the results to puntCat managers but I and the rest of involved people are still working on it Our idea is to use the firmware in our Wireless Communities and maintain it for anyone who needs it leaving the code and compiled images available in our web site http qmp cat 14 CHAPTER 3 QUICK MESH PROJECT CHAPTER 4 Small research in Battle of Mesh 4 1 The Battle of Mesh From the official Wireless Battle of Mesh website The Wireless Battle of the Mesh is an event that aims at bringing together people from across Europe to test the performance of different routing protocols for ad hoc networks like Babel B A T M A N BMX OLSR a
9. conditions because we are using the same channel for all devices the hardware is not powerful we are using five protocols in each node so CPU is always busy and the radio antenna is designed for indoor use then the connection between nodes placed far away is quite bad However this is not a bad point because the protocols are forced to use their best functionalities The tests have been performed from one host to all network hosts as shows the testbed scenario image ping Connected Figure 4 1 Battle of Mesh testbed scenario We will use the following notation to identify the nodes e Main node The central node which performs the pings e Node The rest of the nodes e N Total of nodes seen by Main node From the Main node we are performing the following probes for each Node 1 50 pings with an interval of 0 1ms to get the neighbours ARP this is not stored 2 5 tests of 20 pings with an interval of 0 2ms to destination node 4 3 PROTOCOLS 17 The average latency of the 20 pings for each probe is stored If no reply recived a 0 is stored Then the amount of data collected for each test is SAN We have done a total of 3 tests identified by the start hour 1 T519 05 19 2 T630 06 30 3 T744 07 44 As can be seen every test needs near one hour to finish 4 3 Protocols There are a lot of mesh protocols and a lot of different flavors However in Battle of Mesh are only participating the most active ones wh
10. d 39 19 9 Networks V Lew bU NUR oe NOS oh ee etter te is ERR ERO E 39 13 4 Wireless omui ud a tp eo De aru ode t 40 53 9 Lunnels S Abies d ge ead e d thi atium amp ees 41 7 4 Autoconfiguration system 4 ers 42 Ted Wireless k u S en ae VERTS Goa ae up eds 42 7 42 eNebwOIE cquo eie xk dos vex S an b d GE db eR ERE eee 45 7 5 Management tools 4 s ua nu RR eo nn 46 1 9 Web Interfaces sere lt Pe eo em rede RR C ds 46 59 2 Shell Tools 5 ue 23 22 Mu de xt eae ee te ee enge s Ae e 46 8 Getting the QMP firmware 49 Sub SDR TOE OMP Lolli vm RW BS amp inre rer ee XR RR 49 9 Experience and conclusions 51 Bibliography 53 Glossary 55 CHAPTER 1 Introduction 1 1 Scope Objectives The main purpose of QMP Quick Mesh Project is to provide an open and free software solution for the quick deployment of a WiFi network based on Mesh MANET technology For an accurate explanation look chapter The objective of this document is to collect the principal decisions why and how taken during the development of the project QMP Also it tries to be a starting point for all people who want to become involved in the technical aspects of it However these issues are shown from a general point of view because most of them can be found documented in other places We are only detailing some points which are considered the most important ones dir
11. different roles In such structure all the participants have a voice and the same decision power during the decision making process The director only manages if there is no way decides if there is no way in which the rest of the participants can reach consensus With this paradigm I felt totally involved in the project and I put all my enthusiasm in it Since the work that we are doing is going to be used in our own Network Communities I felt part of it as well I enjoyed being involved in an open and cooperative project I find it a way to enjoy doing what I like and helping other people at the same time Thanks to that I dedicated more hours to the QMP development than the total amount shown in the budget table in section 3 3 I also developed more features than the listed on our project plan just because I found them interesting Referring to the total cost of the project I think it is clear that if it had been developed in a company like environment the price would be much higher In the current study of costs there are a lot of hours invested for free by volunteers mainly from the Director and Responsible who from the beginning decided to do not get any money We are not expecting a profit margin so we are able to sell or rent our hardware for a price close to the original one at a very affordable rates Furthermore we only use free software reducing the expenses considerably too The proof of that is the following example the price of a
12. how clients are connecting to internet 7 2 The interfaces structure Multiple IPs means multiple network interfaces QMP is using five different kind of interfaces each one with his own function They are described in next sections and in the last one an example can be seen 7 2 1 Mesh interfaces For mesh interfaces we use 802 1Q VLAN specified in RFC2815 11 It is needed because QMP system has multiple routing protocols VLAN allow layer 2 isolation so each protocol is using his own tag to send the packets For instance if we configure eth0 as mesh device and we use tags 11 12 and 13 System network interfaces for routing protocols would be the following BMX6 eth0 11 OLSR6 eth0 12 Babel eth0 13 7 2 2 LAN interfaces LAN interfaces are used for final client connections A virtual bridge interface container named br lan will be created with all physical interfaces specified as LAN in configuration A DHCPv4 server and IPv6 stateless auto configuration 12 will be enabled on it 7 2 3 WAN interface WAN interface is used to connect with other networks for instance Internet A DHCPvA client and IPv6 stateless auto configuration requests will be enabled on it 7 2 THE INTERFACES STRUCTURE 37 7 2 4 NIIT 4t06 6to4 interfaces NIIT 15 is an 4in6 tunneling technique developed by the Freifunk community and used in OLSR to transport IPv4 packets over an IPv6 core network It is not a standard but has many similari
13. the following chapters 3 2 FUNDING 11 3 2 Funding Most of this kind of projects are freely contributed people do it just because they believe in it and enjoy working on it However a budget of money is always appreciated because new hardware can be purchased and developers are able to devote more time to it The QMP project draft was sent to a competition organized by puntCat Foundation the cat top level domain managers asking for projects which pushed the social telecommunications To try to satisfy some of those economic needs QMP was presented with the title System to cover quick deployments and it was one of the seven winners budget of fifteen thousand Euros was given to the project for spend in one year according the project details see 3 3 The money obtained from puntCat Foundation covers the development of the software for quick deployments and the purchase of eight new nodes To make it self sustainable after the funded year the idea is to rent the nodes to anyone who need them The price depends on who is asking for it and they might be free for non profit entities One important point to consider is that QMP is more than what was described in the draft sent to the puntCat competition In it the money will only cover the quick deployment but QMP is also as described in 3 1 2 a system for Wireless Communities Funding is not covering this part Figure 3 2 PuntCat foundation logo 3 3 Tasks and bud
14. 10 11 12 13 14 15 16 Adhoc On Demand Distance Vector http www ietf org rfc rfc3561 txt Creating OpenWRT packages http wiki openwrt org doc devel packages Destination Sequenced Distance Vector http www cs virginia edu cl7v cs851 papers dsdv sigcomm94 pdf Draft of BATMAN RFC sent in 2008 to IETF http tools ietf org html draft wunderlich openmesh manet routing 00 Eigennet project http gitorious org eigennet Enhanced Interior Gateway Routing Protocol http www cisco com en US tech tk365 technologies_ tech_note09186a0080093f07 shtm Generic Packet Tunneling in IPv6 http tools ietf org html rfc2473 Getting behind the routing vodoo of BATMAN http www open mesh org wiki batmand RoutingVodoo GNU Make documentation http www gnu org software make IANA Considerations and IETF Protocol Usage for IEEE 802 Parameters http tools ietf org html rfc5342 Integrated Service Mappings on IEEE 802 Networks http tools ietf org html rfc2815 IPv6 stateless autoconfiguration ftp ftp ripe net rfc rfc2462 txt LUCI web interface API documentation http luci subsignal org MANET essential technology for pervasive computing http www mediateam oulu fi publications pdf 92 p NIIT IPv4 unicast traffic through an IPv6 http wiki freifunk net Niit OpenWRT feeds system http wiki openwrt org doc devel feeds 53 54 17 18 19 20 21 22 23 BIBLIOGRAPHY op
15. 4 scope link wlanO 12 link ether 00 80 48 72 d4 XX brd ff ff ff ff ff ff inet6 d02 80 1 280 48ff fe72 d4XX 128 scope global inet6 fe80 280 48ff fe72 d4XX 64 scope link wlan2 link ether 00 80 48 72 e5 XX brd ff ff ff ff ff ff inet6 fe80 280 48ff fe72 e5XX 64 scope link wlan2 11 link ether 00 80 48 72 e5 XX brd ff ff ff ff ff ff inet6 fd02 80 2 280 48ff fe72 e5XX 128 scope global inet6 fe80 280 48ff fe72 e5XX 64 scope link wlan2 12 link ether 00 80 48 72 e5 09 brd ff ff ff ff ff ff inet6 fd02 80 3 280 48ff fe72 e5XX 128 scope global inet6 fe80 280 48ff fe72 e5XX 64 scope link bmx6_in0056 link tunnel6 d02 80 1 280 48ff fe72 d4XX peer d02 80 1 280 48ff fe72 ffff inet6 fe80 215 6dff fec7 9aXX 64 scope link Code 7 1 System interfaces example 7 3 UCI QMP CONFIGURATION FILE 39 7 3 UCI configuration file This chapter describes how the QMP configuration works settings are located in file etc config qmp which is managed by the UCI 20 system The idea is that an end user only needs to modify these options using the management tools like Web Interface or qmpcontrol script According to these values QMP is creating or modifying all needed files and performing the required actions The QMP main configuration file has several sections 7 3 1 Node configuration This is a small section with some settings related to the node config qmp node option community node id 901
16. CN in Europe is Guifi net It was born on a rural village of Catalonia in 2004 Currently Guifi net has more than 15 000 working nodes and 26 000 kilometers of wireless links It is mainly operating in Catalonia but the project is open to anybody from any place around the world d Guifi net Catalonia area map of links in January 2012 As described before decentralized is a structure which fits perfectly in this FWCN environ ment However it is hard to find a software hardware solution for these networks because the companies are normally not interested in it due to the lack of classic business opportunities That is why most part of the development is made by open hacker communities working in Mesh protocols and MANET systems 2 3 2 Mesh and MANET networks In this section we are going to introduce Mesh MANET networks to see them in detail check the external documents 14 18 Strictly speaking a Mesh network is one where all nodes participants are routers meaning that all the nodes accept and forward packets from other nodes according to the routing rules Thanks to this property the physical topology of the network is only restricted by the need of all nodes to be connected through at least one link To be able to use any participant as router a coordinating system is needed This must be flexible and automatic way to ensure the stability and scalability of the network Here is where dynamic routing protocols become important see
17. QMP is using this feature for several things The positioning Map information is transmitted using it There is a chat in web interface which uses it too And in the future we are planning to use it for more purposes like statistics captive portal MAC filter rules etc lJavaScript Object Notation is a lightweight text based open standard designed for human readable data interchange It is derived from the JavaScript scripting language for representing simple data structures and associative arrays called objects CHAPTER OpenWRT a basis for QMP QMP is an operating system firmware for embedded network devices but it is not developed from scratch As part of the free software community QMP is able to benefit from other projects like OpenWRT 6 1 Brief introduction to OpenWRT OpenWRT is a Linux based distribution for embedded devices mainly routers The project started when the Linksys company published the firmware for their WRT54G wireless router under the GPL license Using this code as a basis a community of developers created derivative versions Originally the support was limited to WRT54G series but with the time it was extended to many other chipsets and architectures It was a very important point for non commercial wireless networking because those distributions offers many features no previously found in consumer lever routers Here a list with some of most important features in OpenWRT operating system sourc
18. Quick deployment network using MANET Pau Escrich Garcia January 2012 Acknowledgements Thanks to Leandro Navarro for the global revision of the document Roger Baig for the technical help Axel Neumann for the great job done with BMX6 and the technical support Dario Garcia for the language corrections Anna Juan for the language correction and moral support The QMP team for the great job done during this year The Guifi net project just for being so amazing Contents Introduction 11 1 2 Organization of this document State of art 2 1 2 2 2 3 Centralized and static networks Decentralized and flexible networks The need of decentralized networks 2 3 1 Free Wireless Community Networks 2 3 2 Mesh and MANET networks Quick Mesh Project 3 1 3 2 3 3 3 4 General overview 3 1 1 Quick deployments 3 1 2 Network communities Funding Tasks and budget My contribution Small research in Battle of Mesh 4 1 4 2 4 3 4 4 4 5 The Battle of Mesh Testbed Protocols 4 3 1 Technical details or our testbed Results 3 du RE e 4 4 1 Nodes seen 4 4 2 Successful pings 4 4 3 Latency Conclusions fro the tests 10 10 11 11 13 ii CONTENTS 5 Routing protocol for QMP 23 5 1 Inside BM RG m LARA ELMAR RE ue MU A ES asd 24 5 1 1 Originator Messages 24 5 1 2 Protocol data structure 25 5 1
19. T519 T630 T744 Babel A 89 137 110 OLSR A 221 40 86 BMX6 A 62 131 56 Batman B 45 164 81 Batman adv B 39 37 39 45 CONCLUSIONS FRO THE TESTS 21 250 Babel A Olsr A Bmx6 A 150 Batman B Batman adv B 100 50 0 T519 T630 T744 Figure 4 4 Latency by protocol less is better 4 5 Conclusions fro the tests There is not a clear conclusion about this test and it can just give us a small idea about how the protocols are working More kind of probes are needed to have relevant conclusions for instance a persistent transport protocol like TCP should operate different we are using only ICMP packets Also as can be seen for each test the results are quite different it means that the environment is changing very fast and we are performing tests in serial one after the other for each protocol Another important thing in mesh networks that we are not considering is the mobility our testbed is an static testbed and many of these protocols have implemented interesting features for a mobile environment However some patterns can be shown in the results In nodes seen is clear that BMX6 and Batman adv have the best results The average number of nodes seen percent for each one is 82 and 8996 respectively quite higher than the rest And also the communication between nodes seems to be very stable the average for BMX6 is 9996 and for Batman adv 97 it is deduced from successful ping
20. ame i 10 202 1 1 For QMP we chose a different method we decided to use GNU Make to manage all functions related with SDK We are using a single Makefile with all needed functionalities which helps developers to compile try new features and upload changes Also it can be used by a end user just to generate his own firmware We decided to use this system instead of the old one because the code is cleaner thanks to the powerful programming syntax of the GNU Make tool and to lgit qmp cat qmp git 7http qmp cat qmk firm Shttp gsf guifi net repository 49 50 CHAPTER 8 GETTING THE QMP FIRMWARE win compatibility with OpenWRT since it uses the same method We named this tool QMP Firmware Generator qmpfw and it is available in our git repository The needed steps to compile the system from scratch are the following 1 Download OpenWRT source 2 Download QMP source Integrate both using the feeds system A C Configure parameters according to the target hardware and specific options 5 Compile and leave image ready to be installed To perform all these actions we need to execute the following commands git clone git qmp cat qmpfw git cd qmpfw make 11 build When done we will find the output images ready to be installed in the directory called images As can be seen in the example the destination target is specified using variable T current available targets are e alix PCen
21. attributes like default min max values help syntax parameters This file provides a detailed summary of the current configuration of the daemon descriptions This directory holds one file for each currently active node in the network describing the attributes like IP addresses hostname IDs of each node respectively Volatile information status uptime CPU load globalID number of active nodes version interfaces names status bandwidth IP address etc links outgoing interface to neighbour links neighbour IP addresses link qualities originators list of currently originators in the network primary ip metric etc SMS This plug in uses routing packets to transmit any information from one node to the whole network The good point is that propagation works even if there is no continuous data path Even though the WiFi network is under bad conditions because the Wireless noise distance between nodes etc the data will be propagated However in the current implemen tation there exist a maximum size limit of 300 Bytes for each file The of the sms plug in is very simple It simply clones the content of one or more files given by one node to all other nodes All other nodes can do the same Once started each node will have two directories var run bmz6 sms rcud and var run bmz6 sms rcud Files put into the send folder will be cloned to all other nodes inside rcvd folder
22. chapter 5 for more details 1Stop Online Piracy Act http en wikipedia org wiki Stop Online Piracy Act http www guifi net 2 8 THE NEED OF DECENTRALIZED NETWORKS 7 MANET Mobile Ad Hoc Network is a network with Mesh properties using 802 11 Ad Hoc mode as link layer Thanks to that mobility is allowed nodes can be in movement and frequently switched on off without problems It is clear that when speaking of decentralized and flexible networks we are actually referring to MANET networks It is not the only technology to do so but it is the most widely used and supported CHAPTER 2 STATE OF ART CHAPTER Quick Mesh Project 3 1 General overview The QMP project was drafted by Guifi net active members during year 2010 They did not start from scratch since the idea of this project was based on another project started in 2003 by some networking folks from Gracia district in Barcelona GSF GSF people started working with the first Wifi routers which allowed the installation of some Linux distribution to build their own free outdoor wireless network After a few years this small network joined to Guifi net project which was expanding quite fast within the Catalonia region Unlike most Guifi net parts GSF uses Mesh routing protocols and ad hoc mode as link layer In 2009 a new need appeared organization asked Guifi net members to cover a big public demonstration which took place in Barcelona it was named t
23. d the firmware OpenWRT provides a tool named menuconfig where the user is able to select the packages which will be compiled and wrapped in the final image ready to install in the device as shown in figure 6 1 Figure 6 1 Selecting QMP packages using the menuconfig tool QMP contains among others two main packages described in the next sections IOhttp git scm com 30 CHAPTER 6 OPENWRT A BASIS FOR QMP 6 4 1 qmp small node This package contains a Makefile which includes all mandatory dependencies by QMP and all specific files mainly scripts An example of dependencies could be bmz6 luci ip kmod ath9k So when qmp small node is enabled the rest of needed packages are enabled too It is named small node because it is designed to be used in low performance devices with at least 4MB of internal memory 6 4 2 qmp big node This package is like a meta package just containing dependences It includes qmp small node and a set of packages not mandatory for QMP but useful for the final user an example could be tcpdump openvpn nmap It is designed for high performance devices with at least 8MB of internal memory CHAPTER 7 Inside the qMp development In this chapter the main software components provided by QMP are listed and from a global point of view its operation is described of them are made using the scripting languages Shell Scripting and Lua because both are small fast powerful and OpenWRT c
24. development changing all the time Most of them are born in hacker s environments where the maintainers are doing it just for fun or for their own network Currently one of the most stable DRPs is OLSR probably because it is based on a RFC from 2003 so the implementations are quite mature But it is also a counterpoint because from 2003 to nowadays WiFi networks and devices changed a lot Babel and BATMAN based protocols are newer and designed for Wireless Communities they have interesting features based on the experience of the community people The choice of one DRP for the system means that it will be depending always on it If in some moment the developers decide to stop development or to implement some new feature that makes it incompatible with the network every single node in our system would have to be modified Also if this protocol has some issue which for instance makes some nodes unreachable these nodes would become unreachable and therefore unable to be upgraded by the network administrators According to these points the QMP team decided to use more than one DRP 1 BMX6 as the main DRP 2 OLSR6 as a backup DRP 23 24 CHAPTER 5 ROUTING PROTOCOL FOR QMP 3 Babel as a backup DRP but optional We wanted to have the three flavors in our system BATMAN based protocol OLSR and Babel BMX6 and Batman adv have the best results in the Battle of Mesh evaluation but we chose BMX6 because it has some features very interesting for u
25. e Wikipedia org 23 e writable root file system enabling users to add remove or modify any file e The package manager opkg similar to dpkg or pacman which enables users to install and remove software 17 e package repository containing about 2 000 packages e Sysupgrade preserving selected configuration files on firmware upgrade e UCI unified configuration interface intended to unify and simplify the configuration of the entire system 20 e Extensible configuration of your network involving VLAN e Customizable methods to filter manipulate delay and rearrange network packets 27 28 CHAPTER 6 OPENWRT BASIS FOR QMP An extensive Ajax enabled web interface thanks to the LuCI project 13 User configurable hardware buttons Regular bug fixes and updates even for devices no longer supported by their manufac turers e USB devices support for 3G printer file sharing audio video webcam etc 6 2 Why OpenWRT We decided to use OpenWRT as a base operating system instead other options for embedded devices like VoyageLinux or Tomato Firmware We found several pros in favor of OpenWRT The first one is that most Wireless Communities are using it It is the most widespread operating system most extended in this kind of environment so there exist a huge group of people maintaining it reporting bugs and adding new features In consequence packages are often ported to the repositories and constantly updated Al
26. eader is either 2 or 4 bytes depending on the value of the is_short flag which corresponds to the first byte of the frame header Frame name Description HELLO ADV Hello advertisement Used for letting neighbouring nodes detect the link quality in transmit direction from sending to receiving node RP ADV Rx probe advertisement Used for reporting about reception rate of hello messages from neighbouring nodes OGM ADV OGM advertisement Used for updating periodically route and metric infor mation over the mesh OGM_ACK OGM acknowledgement Used for acknowledging the previously reception of a full OGM_ADV frame The frame data consists of an optional data header and zero or more messages 26 CHAPTER 5 ROUTING PROTOCOL FOR QMP 5 1 3 SMS and JSON plug in BMX6 can be extended using plug ins A plug in is a set of software components that adds specific abilities to an application During the development of QMP two plug ins have been created JSON This plug in exports all available status and topology information in JSON notation using standard file system regular file Data is classified as durable and volatile Files can be found in directory var run bmz6 json It is an interesting feature for QMP we used it to get network status information from the Web Interface Durable information options This file provides a detailed description of all available BMX6 configuration options and its
27. ectly related and owned by QMP The idea of this document is that if does not end here but is extended and improved when system evolves 1 2 Organization of this document The chapter 2 State of art explains the current state of the related points regarding to this project and the need of building a system like QMP It introduces some important concepts that the reader should know before starts with the next chapters Chapter 3 describes the general aspects of QMP how why and where was it born The tasks budget specification and the job I did within the project The two following chapters 4 and 5 are focused in the decision of the routing protocol listing and testing some of them There is a detailed view of the selected one Chapter 6 is referring to another important decision which drove the direction of the devel opment the operating system used as a basis In this chapter are also explained some technical points regarding the integration between our software and this operating system Chapters 7 and 8 are more technical and focused to the people interested in QMP details that could be used to get involved in the project Decisions are also explained so it could be interesting in the research environment as well 1 2 CHAPTER 1 INTRODUCTION Finally in 9 I explain my experience and conclusions of the work done during the development of the project CHAPTER 2 State of art 2 1 Centralized and static networks F
28. elation between nodes available and seen more is better 4 4 2 Successful pings We are only considering the nodes seen One successful ping means at least 1 ping reply from 20 each probe It could be translated like To the total amount of nodes seen for each protocol how stable is the communication with it The numbers are percent the relation between nodes seen and successful pings Protocol T519 T630 T744 Babel A 97 78 88 18 94 00 OLSR A 96 52 97 5 94 29 BMX6 A 98 46 100 98 57 Batman B 94 67 89 47 100 Batman adv B 100 97 04 94 57 20 CHAPTER 4 SMALL RESEARCH IN BATTLE OF MESH Batman adv B Batman B 8 1744 Bmx6 A m T630 WT519 Olsr A Babel A 50 100 150 200 250 300 350 Figure 4 3 Successful pings more is better 4 4 3 Latency Latency is the time delay between the packet send and the reply using ICMP protocol It can give us an idea about how fast is the connexion between nodes We observed that some times the latency is very high something like 2000 or more milliseconds for some nodes placed far away from the Main node So to compare it we only consider the nodes seen by all protocols of a group all of them Because in general the nodes only seen by one or two protocols have a very big latency and consider them would be a disadvantage for the protocols which are able to see those nodes Results are in milliseconds ms Protocol
29. er gateway correction MESH devces lerhn em wlano cese Save O Save amp Apply Figure 7 6 Snapshot of the web interface in the Network tab 7 5 2 Shell tools Shell tools are more focused to be used by the administrator There are two one to configure and another to get information They are also used by the auto configuration system and the web interface Thttp luci subsignal org 7 5 MANAGEMENT TOOLS 4T qmpcontrol This tool is written in Shell Script It is located on etc qmp qmp control sh and imports most of the files placed on etc qmp qmp_wireless sh qmp network sh etc Usage usr bin qmpcontrol function params Available functions offer default gw Offers default gw to the network search default gw Search for a default gw in the network configure wifi Configure and apply current wifi settings configure network Configure and apply current network settings apply_netserver Start stop netserver depending on configuration enable_ns_ppt Enable POE passtrought from NanoStation M2 5 devices Be careful with this qmpinfo This tool is written in Lua It uses the library provided by OpenWRT libiwinfo which gives information about WiFi devices Usage qmpinfo lt question gt options Available questions modes lt dev gt supported modes by the wireless card channels lt dev gt supported channels and supported HT40 type ipv4 print all ipv4 from this compute
30. ere at least one developer is present in the event They are the next with a brief description OLSR Optimized Link State Protocol It is based on the RFC 3626 by IETF We are using the implementation of olsr org website the one used in most Wireless Communities The protocol is link state pro active and uses two different packets HELLO and T C to discover and disseminate the network topology information so it means that each node knows the entire state of the network Babel a loop avoiding distance vector routing protocol One of the youngest protocols is recently defined in RFC 6126 by Juliuz Chroboczek who is nowadays also the main developer It is mainly designed for mobile networks solving problems like routing loops very present in OLSR Link quality is got from IEEE 802 11 MAC so it means a special design also for Wireless networks It is based on the ideas in DSDV 3 AODV 1 and Cisco s EIGRP 6 BATMAN Better Approach To Mobile Adhoc Networking It was born on Freifunk community as a substitute of OLSR with the objective of being easy small and as fast as possible Unlike OLSR it is like Babel a distance vector protocol instead of link state and is using only one kind of messages OGM to discover neighbours So it means that BATMAN is not trying to determinate the whole state of the network each node only knows the first step hop of the entire path to find another node Currently the implementation of BATMAN seems to
31. file GNU Make compatible OpenWRT SDK is providing a set of GNU Make functions to facilitate the implementation of this Makefile To manage feeds the SDK provides an utility located on scripts feeds which permits the easy install remove upgrade of all feed packages For instance the next command would install feeds related with LuCi web interface Ihttp linux voyage hk http www polarcloud com tomato 3https dev openwrt org wiki platforms http uclibc org about html 5http www qca qualcomm com technology technology php nav1 47 6http wiki openwrt org doc devel feeds Thttp www gnu org software make 8http wiki openwrt org doc devel packages 9http luci subsignal org 6 4 QMP MAIN PACKAGES 29 scripts feeds install luci All feed sources are defined in a file named feeds conf which contains the type name and URL For instance this is the line used by the LuCi repository src svn luci http svn luci subsignal org luci branches luci 0 10 contrib package 6 4 QMP main packages QMP uses a feed system as described in chapter 6 3 to include his packets in OpenWRT SDK It uses a git repository located on http qmp cat To use it we should modify feeds conf file and add this entry src git qmp packages git qmp cat qmp git To install it on OpenWRT package tree directory scripts feeds update qmp packages scripts feeds install qmp packages To buil
32. g drip wireless option driver mac80211 option country ES option bssid O27 CALVERT EE BAS e driver Which driver to use mac80211 or madwifi e country Country code to use for WiFi devices e bssid BSSID to use for Ad Hoc devices Specific device section looks like that config wireless option mode adhoc option name qMp option txpower 720 option channel 48 option mac 00 AA BB CC DD EE option device wlanO 7 3 UCI QMP CONFIGURATION FILE 41 or none e mode WiFi supported modes are adhoc ap e name WiFi SSID to use in this device e txpower Transmit power for this device e channel Channel for this device If added 802 11n in mode HT40 will be used e mac MAC identify the device e device Interface name used for this device 7 3 5 Tunnels When a node has connectivity to another network for instance Internet and wants to share it with the rest it must be specified in the tunnels section and also when a node wants to search for access to another network The options might be search offer and ipv4 ipv6 This section is modified in at run time depending on the user needs config qmp tunnels option search ipv4 tunnel 0 0 0 0 0 option offer ipv6 tunnel 08 6 f none used QMP is not configuring the device This is useful if you want to manage it manually or with another sys
33. get This data has been obtained from the original project presented to puntCat competition It details tasks to accomplish and the budget solicited Note that the only expenses are for hired people and hardware Since we are using always free software we do not spent money on this Tasks List of tasks to perform during the project duration They are distributed in four different roles e Director Dir Director of the project his scope is coordinate the group and contact periodically with puntCat e Responsible Res Responsible must help the director with his tasks e Developer Dev In charge of building software and hardware e Documentor Doc In charge of performing needed documentation In each task it is shown the amount of work percent assigned to each role http domini cat 12 CHAPTER 3 QUICK MESH PROJECT Development phase 1 Describe characteristics Dir 7596 Res 25 1 1 Hardware 1 2 Software 1 3 Documentation 2 Hardware selection Dir 2596 Dev 7596 2 1 Choose the hardware to use 3 Software development Dir 1096 Dev 90 3 1 Personalize OpenWRT Linux distribution 3 2 Develop specific functionalities 3 2 1 Auto configuration system 3 2 2 Network integration Mesh MANET 3 2 3 Tools for node management 3 2 4 Tools for internet uplink management 4 Partial tests Dir 2596 Dev 75 4 1 For each node 4 2 Fix detected problems 5 Global tests Dir 25 Dev 75 5 1
34. gines Alix x86 e rspro Ubiquiti RouterStation Pro MIPS ar71xx e rs Ubiquiti RouterStation MIPS ar71xx e nsm5 Ubiquiti NanoStation m5 m2 MIPS ar71xx e fonera FON AP 2X00 MIPS atheros 231x Including a target is quite easy as they are specified in file targets mk included in the repository Multiple targets can be used in the same SDK This is an interesting option for developers because they are able to test all targets using the same Makefile Several extra actions can be executed some of them are listed below check the README file in the software repository to know all of them e make update update sources for all targets e make T TARGET menuconfig executes Ncurses menu config from OpenWRT SDK e make T TARGET kernel menuconfig executes Ncurses menu config from Kernel e make T TARGET clean removes all generated files and leaves the SDK clean e make T TARGET sync_config if a developer makes some change in config files this actions synchronize them with the targets e make config interactive configuration of some QMP parameters CHAPTER Experience and conclusions My professional experience before joining the QMP project was only within private companies It has been my first time in a funded project but not business focused job and the skills that I have acquired have already been really rewarding In this project the organizational structure is completely horizontal although there are several
35. he first 2 0 demonstration In this scenario the most efficient deployment was a flexible network so GSF folks adapted their software to cover it After that more people companies associations and also some political parties were asking for the same kind of deployment to cover their events However GSF software was not designed to be used in this scenario but to be used in a wireless community So some members of GSF and Guifi net decided to start a new project thinking in this new scenario The straight answer to the question what is QMP an operating system firmware for embedded network devices But to understand it maybe it is better to see what it provides e an easy way to deploy a Mesh MANET network e a quick and easy way to extend an Internet uplink to end users QMP has been developed thinking in two different scenarios To achieve quick deployments and to be used in wireless network communities I Gracia Sense Fils http graciasensefils net 10 CHAPTER 3 QUICK MESH PROJECT quick mesh project Figure 3 1 QMP logo 3 1 1 Quick deployments quick deployment means a scenario where a network must be created as fast as possible with the minimum requirements Some examples are concerts meetings demonstrations or natural disasters There are three requirements that we define mandatory for a quick deployment 1 the deployment must be performed as fast as possible 2 it must be able to be done by non technica
36. irst of all must be clarified what we understand when speaking about centralized We are not specifically referring to the physical infrastructure or routing management We are talking about the global structure viewed from a higher abstraction layer Nowadays most of networks are deployed using a centralized topology It means that it exists one node which has the special role to organize the rest Let s call this kind of node Master The other nodes are following his instructions so let s call these other nodes Slaves An example of this is the standard WiFi network used by most of internet end users in their houses figure 2 1 a master a Centralized network with one master centralized network can have multiple Master nodes each one organizing his own slaves Even if master nodes are connected through an horizontal topology the whole network is still being centralized because it is mainly centralized good example is the global network Internet where the Masters Autonomous Systems are connected horizontally speaking BGP But the end clients Slaves are in a second third fourth or more layer knowing nothing about the other layers which ara managed by Masters 4 CHAPTER 2 STATE OF ART b Centralized network with multiple masters centralized network normally is also a static network Static means that nodes at least the master ones are always in the same place doing the same job Each one has his o
37. kg package manager documentation http code google com p opkg Routing Packets into Wireless Mesh Networks http www baumann info public wimob07 pdf Tomato firmware http www polarcloud com tomato UCI database documentation http wiki openwrt org doc uci Unique Local IPv6 Unicast Addresses http tools ietf org html rfc4193 Voyage Linux http linux voyage hk Wikipedia free encyclopedia http wikipedia org Glossary Ad Hoc A decentralized type of wireless network API An application programming interface is a source code based specification Cloud In networking a cloud is a set of nodes connected together deployment Sparse physically a set of nodes due a certain zone firmware Fixed usually rather small programs and or data structures that internally control electronic devices gateway A computer or a network that allows or controls access to another computer or network Hardware A general term for equipment such as wire antennas motherboards handles etc IANA The Internet Assigned Numbers Authority an organisation that oversees IP address Top level domain and Internet protocol code point allocations ICMP Internet Control Message Protocol is one of the core protocols of the Internet Pro tocol Suite IETF The Internet Engineering Task Force develops and promotes Internet standards Internet uplink In a network the point which nodes can use to connect with the Internet network LAN A local area netwo
38. l people 3 it must be possible in most situations To accomplish these points QMP is using the following technologies 1 a mobile ad hoc network MANET with dynamic routing protocol 2 an out of the box system where each node configures itself automatically 3 802 11 WiFi technology so no cables or any other infrastructure is needed For instance to extend an Internet uplink all what people need to do is to spread the nodes and connect at least one to the Internet for all the rest to have access as well 3 1 2 Network communities One of the most important points for a network community is the freedom in this case it means the lack of restrictions Unfortunately that is not totally possible due to the need for instance of hardware and electricity needs and in the current societies people need money to have them But network community users are constantly working to have as less restrictions as possible to build a system which anybody with technical knowledge or not is able to install and use In this regard QMP is providing an out of the box solution easy and free for outdoor Wireless deployments Nowadays there are some communities which have developed their own software and which has been working for several years But there is an important point that will change the current scenarios and will leave some of these systems deprecated the IP protocol version 6 IPv6 QMP provides native IPv6 support as we will see it in
39. late 7 4 2 Network The networking auto configuration system gets and applies the values from the network section of the QMP configuration file see 7 3 It configures the IP addresses as shown in section 7 1 and creates the interfaces as shown in section 7 2 Finally it configures the routing protocols just specifying in which interface have to be enabled and starts them 46 CHAPTER 7 INSIDE THE QMP DEVELOPMENT 7 5 Management tools QMP provides a set of tools to configure manage and monitor the system and the network But they are not going to be explained in detail as the code can be easily read if interested 7 5 1 Web Interface The web interface is based on LUCI which offers a very powerful API written in Lua It is focused to be used by the end user so it tries to be as intuitive and easy as possible The QMP web interface provides the following features e Networking configuration e Wireless configuration e Splash screen configuration for Acess Point clients e Some tools like bandwidth test or ping results e Nodes position Map based on OpenStreetMaps API e Mesh network information neighbours originators etc and configuration Avances nermerbung Quick Mesh Project 1 Interfaces mI m Y mIr Inturfacun LAN devicas ient Tene Q These devices wil be used for user end connection WAN device Ehi Tas deste wil be usec or Internet co ary oth
40. ly IPv6 is an unknown world where nobody actually knows how is it going to be implemented We are trying to follow always RIPE recommendations but probably some ISP are not going to do the same This is a small list of requirements which our schema must accomplish e it must support both IPv4 and IPv6 e it must be configured automatically e multiple IP in the same node must be supported e multiple routing protocols must be supported e it must be the most human readable as possible e it must be the least overlap as possible to avoid IP conflicts e it must permit a good aggregation level to summarize routes between networks QMP is using three different kind of addresses two IPv6 and one IPv4 They will be seen in detail on next sections e internal Mesh routing is based on IPv6 ULA addresses e external internet routing is based on IPv6 IANA addresses e internal and external routing is based on private IPv4 addresses and NAT Figure 7 2 IP schema for qMp network 34 CHAPTER 7 INSIDE THE QMP DEVELOPMENT 7 1 1 ULA addresses ULA means Unified Local Addresses They are a special IPv6 block fc00 7 defined in RFC4193 21 reserved for use in a private environment so they are not routable in the global Internet In a Mesh cloud the Internet connection is a non mandatory option It means that we need some system to be able to deploy a network without using global IPs ULA was developed for this purpose These ULA addresses will
41. nd Static Routing It is a tournament with a social character If you are a mesh networking enthusiast community networking activist or have an interest in mesh networks you might want to check this out The goal of the WirelessBattleMesh events is to set up hands on testbed for each available mesh routing protocol with a standard test procedure for the different mesh networks During the different WBM events similar hardware and software configuration will be used based on the OpenWRT BoardSupportPackage and packages for each protocol implementation The WBM events are also a great opportunity to develop testing tools for PHY MAC radio layers drivers scripts and PHY analyzers Is clear that this event has a lot of points related with QMP In fact one of QMP members was the organizer of the four edition of it in March 2011 There we met with developers of the most important mesh routing protocols in Europe I took the opportunity to perform a small research about the different protocols 4 2 Testbed In our scenario there are a total of 33 nodes routers distributed randomly inside a big complex near 2000 square meters with the following specifications I http battlemesh org 15 16 CHAPTER 4 SMALL RESEARCH IN BATTLE OF MESH e Hardware Fonera 2100 and Fonera 2200 e One 2 4Ghz 802 11bg radio e One Fast Ethernet port e OpenWRT as operating system with all needed software The testbed wireless network is under bad
42. ng These are the steps followed by qmp autoconf we will see them in detail in the next sections 1 Check if control file exist 2 Configure wireless devices 3 Configure networking 4 Restart affected services 5 Create control file 7 4 1 Wireless All wireless functions can be found in the file etc qmp qmp wireless sh Mainly there are two available functions qmp configure wifi initial Leaves all options from QMP configuration file ready to be applied One of the most important points of this function is that all parameters from the configuration file are optional It means than you can leave any of them empty and the QMP system will fill it using the right one or with a default value For instance the user can just configure these parameters config wireless option mode adhoc option name qMp MESH config wireless option mode ap option name gMp AP 7 4 AUTOCONFIGURATION SYSTEM 43 The first WiFi device found will be used as in Ad Hoc mode with SSID qMp MESH and the second one as an access point with SSID qMp AP Then MAC address and device name will be read from the system channel is calculated according the device specifications and T X power is taken from the default values which can be found in the directory etc qmp templates see figure 7 4 Figure 7 4 qmp configure wifi initial flow diagram An interesting function is also wifi get default which return
43. ompatible such as most OpenWRT tools are also built using these languages Figure 7 1 shows the relation between the components of QMP and OpenWRT and which ones are provided to be used by the final user In the next list there is a short description of each of them but in next sections they are described with more detail Provided by OpenWRT e UCI Unified Configuration Interface is mainly a database system intended to centralize the whole configuration of OpenWRT The syntax is very simple and most of OpenWRT components are using it e LUCI Lua UCI Is a the default web interface provided by OpenWRT written in Lua e Admin webint It is LUCI module admin mod full which contains several useful tools to manage the system e WiFi tools A set of userspace tools to manage WiFi devices Some ones standard like iwconfig or iw and some others provided by OpenWRT iwinfo e Networking set of tools to manage networking devices e Bmx6 The routing protocol used by QMP as part of OpenWRT packets Tash NetBSD bin sh ash is a POSIX compliant shell that is much smaller than bash We take advantage of that by making it the shell on the installation root floppy where space is at a premium Lua combines simple procedural syntax with powerful data description constructs based on associative arrays and extensible semantics Lua is dynamically typed runs by interpreting bytecode for a register based virtual machine and has a
44. professional solution like QMP in the telco world is several times highe Aruba Networks is selling his AirMesh router with similar features than the QMP nodes for 5000 USD Speaking about my future work on the project I think I will be involved on it for a long time It is just the beginning Two months ago we got the first stable version but there are several missing features Now we are going to look for other funding sources to keep the development lhttp www securewirelessworks com Aruba MSRA000 asp 51 52 CHAPTER 9 EXPERIENCE AND CONCLUSIONS active Actually we are already involved in a FP7 and preparing a new project to present to the NLnet foundation Regarding the technical conclusions there is not much left relevant to say because it does not fit into the scope of this document We have made some technical errors during development but that is understandable since we are not a business and we can roll back mistakes without many complications In addition we are always doing research and trying to improve the system because most of the technology we are using is not tested yet or at times not even implemented In this scenario making mistakes should be considered as a good way for improvement as long as something is learned from them since the worse part of those mistakes can be easily solved in our organizational model http cordis europa eu fp7 home en html 3http nlnet nl Bibliography
45. program and related data that provides the instructions for telling a computer what to do and how to do it SSID A wireless network identifier TCP Transmission Control Protocol one of the core protocols of the Internet protocol suite Testbed A platform for experimentation topology Is the layout pattern of interconnections of the various elements links nodes etc of a computer UDP User Datagram Protocol a simple transport protocol used on the Internet WAN A wide area network s a telecommunication network that covers a broad area
46. r excluding localhost hostname print device hostname bwtest lt ipv6 gt perform a bandwidth test nodes list all nodes on MESH network key print qmp key 48 CHAPTER 7 INSIDE THE QMP DEVELOPMENT CHAPTER Getting the QMP firmware As referenced several times in this document QMP source is available in a git repository The Master branch is the current stable one and the rest are testing branches used by developers Following the OpenWRT policy we are using alcoholic drink names to identify the stable releases with the first one being called Ratafia Compiled binaries of the current release can be found here 8 1 SDK for QMP QMP which involves OpenWRT too is a complex system where all needed packets are cross compiled to be used in embedded devices As a result a set of special packets patches scripts etc are needed Providing a SDK Software Development Kit is an important point to main tain and integrate new features in the system As explained in chapter 6 OpenWRT provides a powerful kit In this regard QMP SDK is an abstraction layer of it making it more simple and including new features The predecessor firmware GSF see chapter 3 1 used a shell script as SDK It was user tool including options to build update firmware For instance to generate a binary image for Alix board with name NodeName and IP 10 202 1 1 see for more info we would run the command fwgen d w alix n NodeN
47. r environment like WiFi 2 In a free network to ensure the freedom of the network Those two scenarios fit perfectly in a environment which in last years has experimented a significant growth in the telecommunications world the Free Wireless Community Networks shown in chapter 2 3 1 2 3 1 Free Wireless Community Networks community network is a network made and maintained by the same participants Unlike the model used by the global telecommunication companies which are business focused each user is owner of his stretch following the philosophy make it yourself Using some agreements and organizations e g web site they are able to connect with neighbours neighbours of the neighbours and so on These communities are normally open and free in the sense explained in section 2 3 Most of them are using WiFi technology because it is the easiest and cheapest way to deploy an outdoor network This model named Free Wireless Community Network FWCN is nowadays being used in many places and is growing every day Probably one of the main reasons of this growth is the governments policy which day by day tries to control the Internet and the connections within it An example of this is the cut of communications made in Egypt during the 2011 revolution 6 CHAPTER 2 STATE OF ART or the SOPA bill currently being discussed in the USA A community network is owned by the users the government is not able to control it One of the biggest FW
48. rk LAN is a computer network that interconnects computers in a limited area Layer OSI model It is a prescription of characterizing and standardizing the functions of a communications system in terms of abstraction layers Makefile Main file used to specify the operation that make command must do meta package Is a special package that only contains metadata 55 96 GLOSSARY NAT Network address translation is the process of modifying IP address information in IP packet headers neighbour node is neighbour of another node if they are directly connected Network community set of nodes placed by the people to connect between them node An active device attached to a computer network or other telecommunications net work out of the box A property that means it does not need manual configuration Ping Administration utility used to test the reachability of a host on an Internet Protocol IP RFC Request for Comments is a memorandum published by the IETF describing methods behaviors research or innovations RIPE R seaux IP Europ ens is a forum open to all parties with an interest in the technical development of the Internet router device that forwards data packets between computer networks SDK A software development kit is typically a set of software development tools that allows for the creation of applications shell piece of software that provides an interface for users of an operating system Software A computer
49. rnet 2 Search for Internet 3 Announces the IP range available 4 Read the available range 5 Announces the minimum amount of IPs that can give 6 Ask for the range and the amount of IPs 7 Using unicast AHCP the gateway node gives the desired IP range to the client After that the node starts offering stateless autoconfiguration addresses to the clients 7 1 3 IPv4 addresses IPv4 addresses are needed to connect to the current global Internet normally using private addresses trough a NAT address translation We need one IP for each routing protocol because then we are able to choose which protocol to use when we are connecting to some other node just choosing the right IP These IPs will be configured in a non physical interface because they are used over an IPv6 tunnelled interface When defining the IPs we have three options to choose depending on our scenario if quick deployment or wireless community Selecting exactly the IPs that we want to use BMX6 OLSR6 Babel 10 140 48 1 24 10 140 49 1 24 10 140 50 1 24 Selecting 24 range so the lasts 8 bit will be calculated randomly BMX6 OLSR6 Babel 10 140 48 10 140 49 10 140 50 Selecting a 16 range so the lasts 16 bit will be calculated randomly BMX6 OLSR6 Babel 10 201 10 202 10 203 36 CHAPTER 7 INSIDE THE QMP DEVELOPMENT Clients get IP using auto configuration IPv4 NATed Clients get IP using DHCPv4 Figure 7 3 Schema about
50. s default values Taking a look on the flow diagram 7 5 we can see for instance how the channel is calculated to try to avoid channel collisions that cause noise In next table we can see the default selection of the channel depending on the interface name and mode considering three WiFi devices compatible with the 2 4GHz band adhoc ap wlanO 11 1 wlanl 9 3 wlan2 6 5 qmp configure wifi This function applies WiFi parameters from the QMP configuration file to the OpenWRT system To do that it uses a set of files placed on etc qmp templates They have a special name that indicates on which case it must be used wireless lt driver gt lt mode gt n For instance wireless mac80211 adhoc n n means 802 11n mode or wireless madwifi ap These files are UCI 20 friendly and contain some special words starting with QMP that will be used to substitute by specific ones during function execution time wireless QMP_DEVICE wifi device wireless QMP_DEVICE type mac80211 wireless QMP_DEVICE macaddr QMP_MAC wireless QMP_DEVICE channel QMP_CHANNEL 44 CHAPTER 7 INSIDE THE QMP DEVELOPMENT what parameter to configure device Wi Fi device used Yes mode device mode adhoc ap Figure 7 5 wifi get default flow diagram 7 4 AUTOCONFIGURATION SYSTEM 45 wireless Qwifi iface tQMP INDEX mode adhoc wireless wifi iface QMP_INDEX bssid QMP_BSSID Code 7 3 example of wireless temp
51. s is the bridge between all LAN devices niit4to6 Tunneled interface using NIIT It contains all IPv4 addresses used inside the Mesh network and IPv6 IANA if present bmx6_in0056 This is the tunnel created by BMX6 to share the Internet connection with the rest of the network http freifunk net 5http gitorious org eigennet 38 CHAPTER 7 INSIDE THE QMP DEVELOPMENT etho link ether 00 15 6d c7 9a XX brd ff ff ff ff ff ff inet 192 168 1 20 24 brd 192 168 1 255 scope global ethO inet6 fd02 0000 0080 78f9 215 6dff 64 scope global inet6 fe80 215 6dff fec7 9a5d 64 scope link 1 link ether 00 15 6d c7 9a XX brd ff ff ff ff ff ff wlani link ether 00 80 48 72 d6 XX brd ff ff ff ff ff ff br lan Wink ether 00 15 6de 9a XX brd fie thet fect ff Ef inet 172 30 22 1 16 brd 172 30 255 255 scope global br lan inet6 fe80 215 6dff fec7 9a5e 64 scope link niit4to6 link ether 36 78 fc 1c a9 XX brd ff ff ff ff ff ff inet 10 201 0 124 32 brd 10 201 0 124 scope global niit4to6 1 inet 10 202 0 124 32 brd 10 202 0 124 scope global niit4to6 2 inet6 2012 0 0 7c 1 64 scope global inet6 2011 0 0 7c 1 64 scope global inet6 fe80 3478 fcff feic a94e 64 scope link wlanO link ether 00 80 48 72 d4 XX brd ff ff ff ff ff ff inet6 fe80 280 48ff fe72 d4XX 64 scope link wlanO 11 link ether 00 80 48 72 d4 XX brd ff ff ff ff ff ff inet6 d02 80 0 280 48ff fe72 d4XX 128 scope global inet6 fe80 280 48ff fe72 d4XX 6
52. s not implemented in Batman adv n example is the SMS plugin or the smart gateway selection 5 1 Inside BMX6 As explained before we chose BMX6 as main protocol so in this section it will be explained with more detail Documentation of BMX6 is very sparse I got the next information with help from Axel Neumann main developer and Roger Baig working in a research thesis about mesh protocols As starting point the two main points which characterize BMX6 e Pro active As opposite of reactive it maintains an updated routing table by periodically distributing information through the network Thus it can be considered loop free e Destination sequenced Distance vector DSDV Table driven routing scheme where the node does not know the entire network topology just the best neighbour for a specific destination See 3 for more details The following sections describe some specific features of BMX6 and how they work 5 1 1 Originator Messages As its predecessor BATMAN BMX6 uses periodic packets named OriGinator Message OGM These are UDP User Datagram Packet sent normally every half second to the direct connected neighbours using a broadcast address as destination OGM are used to update routes and metrics over the mesh network When a node receives this message from other node it is forwarded through all interfaces So thanks to this UDP flooding a node is able to know the existence of all mesh participants see figure 5 1
53. s probe We have to consider than the scenario is very complicated for communication this good numbers are a proof that these mesh protocols have a good implementation and they can operate in a very bad conditions Also do not forget about OLSR his numbers are close to BMX6 ones 22 CHAPTER 4 SMALL RESEARCH IN BATTLE OF MESH CHAPTER 5 Routing protocol for QMP The routing protocol is probably the most important component of a Mesh network When managing a small number of nodes any kind of routing protocol may work fine also static but as the network grows the routing becomes key for the final performance of the net Currently most production networks use protocols like BGP or OSPF which is probably a good choice for static an environments mainly because they are quite simple and stable However as described in chapter 3 our scenario is completely different the network we work with is changing very fast and static protocols are not developed for such cases It is necessary a more complex protocol For instance other metrics beyond number of hops should be taken into account such the packet losses path saturation link quality and so on In chapter 4 there is small evaluation about the different routing protocols designed for Mesh networks which we will call dynamic routing protocols or DRP At this point we should answer the question which DRP should be used in QMP it is hard to say as all these DRP are in constant
54. se the group B is not using Ethernet interface then according our scenario the total number of nodes that can be seen by group B is 25 instead of 33 seen by group A The command used to perform pings is the next ping c 20 i 0 2 Host q W1 s64 4 4 Results 4 4 1 Nodes seen There are 33 nodes for group A and 25 for group B However not all of them are always reachable maybe they appear in the description but the connection is not possible because the path is impossible or they are not using the best one To probe that we are performing 50 pings for each node if there are not at least one reply we consider that this nodes is not reachable from the Main node In the next table we can see the total amount of nodes seen for each protocol Protocol T519 T630 T744 Babel A 18 22 20 OLSR A 23 24 28 BMX6 A 26 27 28 Batman B 15 19 18 Batman adv B 24 20 23 Remember that there are two different groups and B so to compare all protocols we should use the relation between available and seen nodes percent as shown in the next table 44 RESULTS 19 Protocol Available nodes T519 T630 744 Babel A 33 5596 67 61 OLSR 33 70 73 85 BMX6 A 33 79 82 85 Batman 25 60 76 72 Batman adv B 25 96 80 9296 Batman adv B Batman B 8 T744 Bmx6 A mT630 mT519 Olsr A Babel A 50 100 150 200 250 300 e Figure 4 2 R
55. so the powerful SDK software development kid provided by OpenWRT permits to cross compile the sources for many different architectures like x86 MIPS or ARM complete list of supported architectures can be found here That is important because most of routers are using these kind of small architectures not x86 To be able to run properly in these low performance computers the system is very optimized Binaries are normally very small because they are compiled using uClibc much smaller than GNU libc and stripped from unnecessary functions Also the applications are cautiously selected e g busybox instead all common shell tools and some of them even modified to be smaller For instance hostapd mini which is a hostapd version with only WPA PSK support most common used security protocol in WiFi networks And as last point the wireless device support in OpenWRT is probably the best one of all Linux based distribution especially in Atheros devices where some developers are also involved directly in the operating system An example of the last year 2011 OpenWRT was the first system which supported IBSS Ad Hoc Wifi mode in 802 11n and HT40 MiMo 2x20MHz channels for mac80211 devices most common 6 3 QMP as part of OpenWRT To integrate QMP with OpenWRT we use the feeds system A feed is a collection of packages sharing a common location using a remote server or a local filesystem package is a directory which contains a Make
56. standard 10 For example this is how the block is split for this IP fcba beff f012 0001 0000 82b1 5981 a1cb ULA prefix QMP prefix Cloud X Interface MAC fd b0b0b0b0 12 000 1 000082b15981alcb Considering the same node these four IPs would be assigned to the different interfaces according to the descriptions Note that AP interfaces are using 64 this is because we are going to use IPv6 auto configuration feature for the clients In bold the changing parts e BMX6 and ethO used for MESH fcb0 b0b0 b000 0000 MAC 128 e BMX6 and ethl used for MESH fcb0 b0b0 b000 0001 MAC 128 e OLSR6 and ethO used for MESH fcb0 b0b0 b100 0000 MAC 128 e br lan used for clients fcb0 b0b0 b000 9999 64 7 1 THE IP STRUCTURE 35 7 1 2 IANA addresses We understand IANA addresses as the IP addresses assigned by IANA in fact by RIPE in Europe APNIC in Asia and so on to our ISP and our ISP to us so they are routable to the Internet These are created when a node connects to Internet and share it with the whole network However the addresses got by the Internet node must be at least a 60 because each node needs at least a 64 to provide access to their clients All LAN interfaces inside the same node are bridged so one 64 address is enough to offer connectivity to all connected clients These are the steps to follow in a high abstraction layer Gateway node Client node 1 Announces Inte
57. tem 42 CHAPTER 7 INSIDE THE QMP DEVELOPMENT 7 4 Autoconfiguration system This is one of the most important parts of the QMP firmware because the proper working of it means that user does not need to have technical knowledge He just needs to install the firmware in the device maybe configure some parameters That is enough to be part of the network However each wireless cloud has its own specific options for instance as seen in the IP structure section 7 1 the cloud number or QMP prefix for IPv6 configuration The idea is that network administrators build an image with the specific parameters and leave it available for download for the normal users However if this task is not performed default settings should work too The auto configuration system is executed only the first time the device boots There is a special file placed on gmp_autoconf lets call it control file and a init script on etc init d qmp autoconf This script is executed at each boot but it only operates if the control file does not exist In the first boot the init script will execute all needed functions to leave the system ready and after that it will create the control file So this functions will not be called anymore unless the user or the administrator decides to do it by removing the control file if CONTROL FILE then configure else echo Remove CONTROL_FILE to force reconfiguration fi Code 7 2 Control file checki
58. ties to 4in6 standard RFC2473 7 NIIT is much more simple compared to 4in6 and also can be easily autoconfigured QMP is an IPv6 native network so a tunneling transport layer is needed to use IPv4 The eigennet project from Pisa has already integrated NIIT in their mesh firmware We are using NIIT implementation from their repository 7 2 5 BMX6 tunnel interfaces When a gateway gate to another network is published by some node the rest might be interested in it In this case to use it the interested nodes must request a tunnel between it and the gateway This is due to the kind of source address If this is pointed to some external address like a public Internet address each node in the path between both nodes will process the packets in different way So a direct connection from interested node to gateway node is needed BMX6 implements his own tunnel system creating virtual interfaces with names like bmx6 in0056 or bmx6_out0056 This kind of interfaces are created destroyed at run time 7 2 6 An example of interfaces Figure 7 2 6 an example of QMP interfaces configured in a production node The text has been summarized to facilitate the reading The following table briefly describes the usage of each one eth0 WAN device connected to Internet ethl wlanl LAN devices used to end user connection wlan0 wlan2 Used as Mesh The VLAN interfaces tagged with 11 12 is where routing protocols are running br lan Thi
59. utomatic memory management with incremental garbage collection making it ideal for configuration scripting and rapid prototyping 31 32 CHAPTER 7 INSIDE THE QMP DEVELOPMENT Provided by QMP qmpinfo It is a tool written in Lua to get some information about the QMP system qmpcontrol It is a tool written as a shell script to manage the QMP system qMp webint A LUCI web interface module to manage and monitor QMP system map map tool named b6m and written in Shell Scripting using OpenStreetMaps APT to see the position of all network nodes and how are they connected autoconf system to auto configure all settings So the user does not need to configure anything internet detect A tool named gwck written in Shell Script which is monitoring if the node is connected to internet in this case it modifies all needed things to share it with the whole network QMP OpenWRT Figure 7 1 Relation between QMP and OpenWRT components 3http www openstreetmap org 7 1 THE IP STRUCTURE 33 7 1 The IP structure The IP schema is very important point because it is key for the network scalability and compatibility with other networks It has been discussed a lot of times in our team First decision is to use IPv6 as backbone and IPv4 tunneled over IPv6 Then when finally IPv4 disappears removing it will not have big complications for the network administrators However it has been a challenge for us because current
60. w as WiFi Unlike most of cabled networks WiFi networks are continuously changing The physical infrastructure is not dedicated to data transmission consequently any external agent can change its properties a building a tree the rain or even a bird If the network administrator is giving all the control to one or more nodes Masters and some external agent changes the topology affecting one of these nodes the performance of the network will be affected That is why a non centralized structure could be useful in an outdoor WiFi environment See figure 2 1 c Another important point to take into account is the freedom Freedom in networking means mainly two things First one the lack of an artificial restriction to participate in the network Second one the network is content neutral ergo nobody can prioritize discriminate traffic ac cording the data content or the user owner of this data In a Master Slave structure even if 2 8 THE NEED OF DECENTRALIZED NETWORKS 5 Centralized network Non centralized network mE 5 a y c Centralized VS Decentralized in a outdoor WiFi environment the network is declared free the freedom can be easy violated Since even if it is controlled by some few nodes which have the power to do so in this kind of environment it is important to implement a non centralized structure Two scenarios have been exposed where decentralized networks might be required 1 In a changing link laye
61. wn specific configuration so the deployment of it must be thought in advance This kind of networks are a good solution for a long term system in a non changing environment 2 2 Decentralized and flexible networks A decentralized or non centralized network is the one where all nodes have the same role Neither master nor slaves just participants A very explicit example is the neural network of our brain where all neurons are in the same level broadcasting information to their neighbours However when speaking about centralized we are referring to structure In a horizontal network can also appear a centralized behavior product of for instance an internet uplink If all nodes from the network want to access internet using the same uplink node it will operate like a Master Anyway the network keeps being decentralized because of the structure and not the internal operation But these cases should not be welcome by the network users administrators because they are losing part of the benefits offered by this kind of structure Flexible means as opposed of static that the network can be frequently physically changed and the proper functioning is not affected Existing nodes can move and new ones can be added without the need of changing anything This property is normally linked to a decentralized structure 2 3 The need of decentralized networks Decentralized structures can be useful in networks based on IEEE 802 11 standards commonly kno

Download Pdf Manuals

Related Search

Related Contents

    Geemarc CL200+ Telephone User Manual  

Copyright © All rights reserved. Failed to retrieve file