Mobile Metropolitan Ad hoc Networks

We have approached the problem of location and discovery service. The main goal was to provide a tool able give detailed information about what is available on the network and to give facilities to spread information about what you want to make available. The application is based on the UDI architecture retailed for mobile environment (UDDIm). It has been developed on two main platforms: traditional peer-to-peer (pastry) and integrated with cross layer (sedigned and developed in the project from other partners). The target device are light device (like pda) or not (i.e. laptop). The system is based on free software and it does not need any special component a part from software developed I the project.

The area of ad hoc networking is of long-term nature. However, our results show good potentialities from the innovation and economic standpoint. Specifically, We have identified new applications that can leverage the ad hoc technology to provide valuable services to the user. The city cab scenario (i.e., the use of 802.11 ad hoc networks to replace the currently used taxi radio dispatch systems) is the most promising one. One of the most important components of a city taxi company s operations is its dispatch unit, which informs individual taxis about passenger pickups, assigns passengers to empty taxis, and passes on additional information such as directions and news about weather or traffic conditions. Traditionally, taxi companies have relied on radio dispatchers for these tasks. Radio dispatch systems have variable QoS issues, and there are significant costs associated with installing, running and maintaining them. In many cases, radio licenses need to be obtained to operate these systems, adding considerable cost and creating barriers to entry for small companies. In MobileMAN we proved that a MANET based dispatch system used for communications in a city cab scenario is viable both economically and technically.

The VoIP package contains two main modules; signalling module and data transport module. The signaling module consists of the software component that will initiate the VoIP session with other peer nodes in the Ad Hoc network. This module has been implemented specifically for the Ad Hoc framework since existing implementations did require excessive resources (e.g. CPU, memory, etc). The signalling module implements the SIP signalling protocol and utilizes IP addresses for finding the peer nodes to initiate the VoIP session. The SIP signalling protocol can run on UDP or TCP protocol but in order to minimize the requirements for maintaining the session state in the nodes, the existing implementation uses UDP as the only transport protocol. The session initiation also requires negotiating the media parameters using SDP protocol. In order to minimize the negotiation process the signalling module uses the same codec for the VoIP session (i.e. GSM). Therefore, the signalling module is compliant with the SIP protocol but having a single codec optimises the session set-up. The SIP module is implemented specifically for the Ad Hoc network but the GSM codec is obtained from public source. The data transport module consists of the software component that after the VoIP session is set up, takes care of exchanging the voice packets coded with the selected media format (i.e. in this case GSM is the only codec used in the session). The data transport module implements a RTP client for exchanging the voice packets. The RTP client implements the functions for obtaining the audio samples from the microphone, encoding them using the selected codec (i.e. GSM) and then exchange the packets using the RTP protocol. The RTP client uses an publicly available RTP library for managing the RTP messages. The effectiveness of the developed package has been tested on multi-hop ad hoc networks of up to 3 hops. In the experimentation we used both OLSR and AODV.

The Ad Hoc routing framework is a software package, which can support different Ad Hoc networks routing protocols, such as proactive, reactive and also some hybrid solutions. The Ad Hoc routing framework can be installed in a node (PDA or Laptop) that runs Linux Operating System. The framework has been designed in separate components with clearly defined interfaces. This allows an easy integration of these components and the possibility of adding new functionalities (new routing protocols and other functionalities, such as naming and service discovery). The framework provides general functionalities for both proactive and reactive routing protocols. The existing framework includes a reactive protocol (e.g. AODV) and a proactive protocol (e.g. OSLR). In order to test the framework implementation with constrained devices, in addition to laptops the framework is integrated into a small number of Personal Digital Assistants (PDA) nodes (iPAQ). The iPAQ is a mobile node where the operating system is has been changed to Linux. The Ad Hoc Framework consists of four subsystems: the Common Cache Registry Server, the Reactive modules, the Proactive modules and the Hybrid modules. The Common Cache includes all the modules that must be kept constantly running in the node since they store routing information and other data used by the other modules. The Reactive modules consist of the software modules that implement the reactive routing protocols (e.g. in this case the Reactive module consists of the module that implements AODV). The proactive modules consist of the software modules that implement proactive routing protocols (e.g. in the actual framework the proactive modules contain only a software module that implements OLSR). Finally, the hybrid modules include all modules that will implement hybrid routing protocol such as ZRP. The four modules of the Ad Hoc framework consist of independent software components that implement specific routing protocols and store routing information into a single cache.

By integrating of off-the-shelf HW and SW components with software modules we developed in the project, we obtained two prototypes of a campus-wide MobileMAN. One prototype implements a legacy (layered) TCP/IP architecture. The other prototype implements a cross-layer architecture. On both prototype we integrated a Whiteboard and a UDDI4m application. By exploiting these prototypes, we performed extensive experimental evaluations that provided relevant indications on the performance of current ad hoc networking solutions. Specifically, - We compared proactive (OLSR) and reactive (AODV) routing protocols on realistic testbeds; - We performed a measurement-based analysis of Pastry and CrossROAD performance when running on mobile ad hoc networks; - We experimented a medium size (up to 23 nodes with paths made up of up top 8 hops) ad hoc network implementing the architectures 1 and 2. Currently, this is one of the largest ad hoc testbeds implemented in worldwide research projects.

Communication systems based on self-organizing entities that build up the network without the need or reliance for a pre-established infrastructure represent a challenging scenario that will play an important role in society and economy by providing opportunities for the creation of ad hoc networks and services. However, in order for these services to be successful, they must rely on a network that is secure. The increased sensibility of mobile ad hoc networks (MANETs) with respect to dedicated networks like the Internet derives from the lack of nodes with a predefined role that are responsible for the network operation. Initially, applications of ad hoc networks have been envisioned mainly for crisis situations (e.g., in battlefields or in rescue operations). In these applications, all the nodes of the network belong to a single authority (e.g., a single military unit or a rescue team) and have a common goal. However, the deployment of ad hoc networks for civilian applications has become realistic. In these networks, nodes typically do not belong to the same organizational structure nor to a single authority and they do not pursue a common goal. The lack of an a-priori trust relationship between the members of the network renders security an essential component to enable a realistic deployment and utilization of such open networks. In this project we addressed the security issues raised by open ad hoc networks. We first investigated on the impact of several threats that have been often neglected by the research community when designing ad hoc routing protocols in which all participants were inherently trusted. If these threats can be considered in line with the experience gathered through the study of a variety of attacks on routing protocols for classical networks, in this thesis we point out and analyse a new type of misbehaviour that we called node selfishness, specific to the highly cooperative environment offered by the ad hoc networking paradigm. A simulation-based analysis conducted in our laboratories revealed that node cooperation is essential because unlike networks using dedicated nodes to support basic functions like packet forwarding, routing, and network management, in ad hoc networks those functions are carried out by all available nodes. However, there is no reason to assume that nodes will participate in the network operation, especially in battery powered environment such as a MANET. We proposed a state of the art of basic security services for ad hoc networks, that range from secure routing protocols to key-management services and we focus on various type of cooperation enforcement mechanisms available in the literature. Our research pointed out two challenging research directions that we further investigated in the reminder of the project: the novelty of cooperation requirements and the difficult task of providing security associations without the support of an external infrastructure. We thus propose a cooperation enforcement mechanism based on an original reputation system that we called CORE. We then propose a detailed validation of the CORE mechanism using two different methodologies: in the first method, we use a classical network simulation tool to infer the basic properties of CORE. We then extend our work in order to cope with a sophisticated model of node selfishness that takes into account end-users' "rationality" when operating the nodes. Our validation study shows that CORE is an effective and robust mechanism that stimulates cooperation of selfish nodes; furthermore, through the evaluation of nodes' energy consumption, we provide evidence that CORE also offers incentives for legitimate nodes to use it as a cooperation mechanism in that they can save a non-negligible amount of energy. We conclude the project by proposing a novel authentication scheme for mobile ad hoc networks that does not rely on a pre-established network infrastructure and that does not require any shared organization between the users of the network. In our scheme, a lightweight bootstrap phase is necessary for a node entering the network: by contacting an authentication server a node is able to locally generate authentication credentials that are globally verifiable in the network without the need for the distribution of public key certificates. We also propose a practical utilization of our scheme as a complementary key distribution scheme that enables authentication services demanded by secure routing protocols available in the literature.

CrossROAD represents an optimised p2p system for ad hoc networks, based on the Pastry overlay network model. Specifically it exploits a cross-layer architecture, using network routing table information in order to maintain a correspondence between the physical network topology and the logical address space, where nodes and data are mapped. In order to have a complete and updated knowledge of the network topology, a proactive routing protocol is needed, and for this reason we selected an open source implementation of OLSR (Unik-OLSR v.0.4.8) that allows the definition of libraries dynamically loaded by the routing daemon at the startup, in order to export routing information to other applications, or to define additional information to be sent on the network through the proactive flooding of routing packets. These libraries are called plugins. In our prototype a plugin, called XL-plugin, has been defined in order to encapsulate additional information in routing packets. This information is represented by services identifiers, used to associate to each node the list of services locally provided. When OLSR receives a routing message containing this additional information, it passes the contents to XL-plugin that provides to store services identifiers of other nodes in its local data structures. For this reason XL-plugin maintains two local data structures: LocalService Table and GlobalService Table. Specifically, the LocalService Table maintains the list of services provided by the local node, while the GlobalService Table maintains, for each service present in the network and currently running on CrossROAD, the list of nodes providing it. All entries are timed out in order to preserve the consistency of the service information. In this way, when a node starts running an application on top of CrossROAD, it declares its service identifier and CrossROAD directly establishes a local connection to the plugin in order to receive the list of nodes taking part to that specific overlay. Then, when the local application sends a message with a specified key value, CrossROAD first checks the consistency of its internal data structures with the list provided by the plugin, then it determines the best destination for that key and directly sends the message to it. More details on software architecture of CrossROAD and XL-plugin can be found in deliverable D13.

We developed a complete study of 802.11 use in multi hop ad hoc networks based on extensive experimental measurements and simulations. These studies produced an in depth understanding of 802.11 behaviour in ad hoc mode that was exploited to develop an accurate channel model. These results constitute the basis for the design, implementation and validation of an enhanced 802.11 MAC card that provides significant performance gains with respect to the standard 802.11 cards. The enhanced card has been designed to interoperate with legacy standard cards. More precisely, we achieved the results listed below, and expensively documented in Deliverables D12, D13 and D16. -Extensive measurements to characterize the behaviour of IEEE 802.11 ad hoc networks from which we extracted a model to characterize it. -Design of an enhanced MAC protocol (AOB) for ad hoc networks. The new MAC protocol is compatible with the IEEE 802.11 and provides a better channel utilization and fairness. - Extension of the AOB mechanism, with a credit-based mechanism, to effectively operate in a heterogeneous environment where enhanced and legacy 802.11 cards co-exist. The credit mechanism provides a formal basis to the activities of TG 802.11n that is working toward higher throughput for 802.11 networks. Indeed, AOB extended with the credit mechanism provides an optimised and efficient solution to the multiple transmissions approach currently under study in TGn. The credit mechanism provides an efficient solution to fix 802.11 unfairness in multi-hop scenarios. - Hardware implementation of the enhanced MAC card implementing both the AOB mechanism and its credit-based extension. - Experimental validation of the AOB mechanism on a 4-node testbed. - Implementation in the NS-2 tool of the simulation model of the enhanced MAC card. - Validation of the AOB mechanism via simulation on large-scale networks.

One of the scientific results of the project has been the design of a new architecture for mobile multi-hop ad hoc networks. The developed architecture is based on the introduction of the cross-layer principle in mobile-ad-hoc-network organization still maintaining a layered organization of the network architecture and the compliance with TCP/IP protocol stack. More precisely, the MobileMAN cross-layer architecture is based on a layered organization, which can be enhanced with cross layering interactions if information gathered at different layers of the network stack is shared in a common local memory structure (Network Status, referred to as NeSt). Specifically, - The NeSt, which is the key of the cross-layer architecture, has been completely defined by specifying the NeSt interaction models, and its exported interfaces. - We designed the software architecture of a NeSt prototype supporting cross-layer interactions between a proactive routing protocol and the middleware platform, CrossROAD, which has been developed during the project. - Starting from the software architecture, we implemented a proof-of-concept prototype of our cross-layer architecture. - An experimental phase provided a proof of the benefits of this new approach in MANET design. - A special attention was devoted to study the performance of our cross-layer architecture. To this end we extended the Network Simulator NS-2with a cross-layer interface (XL-interface) that standardizes vertical interactions among protocols according to the MobileMAN cross-layer architecture.

P2p systems provide an effective framework to realize decentralized and scalable data sharing among large sets of users. Furthermore the decentralized and peer-to-peer computing model is suitable for mobile ad hoc networks. However, in the project, using both simulation and experimentation, we showed that two well-known platforms, like Gnutella (un-structured) and Pastry (structured) exhibit poor performance in ad hoc environments due to low tolerance to dynamics such as nodes mobility. To cope with these performance problems, we designed cross-layer versions of Pastry and Gnutella. The cross-layer version of Pastry (CrossROAD) and Gnutella (XL-Gnutella) exploit cross-layer interactions between a proactive routing protocol (OLSR) and the middleware platform mediated via the Network Status (NeSt). Extensive performance evaluation studies (either via simulation or measurements) have shown that our cross-layer solutions provide significant performance gains when a middleware platform is implemented on top of an ad hoc network. More precisely cross-layer solutions are more adaptive and tolerant to ad hoc network dynamics such as nodes mobility and wireless link instabilities.

Livrables

Partager cette page

Télécharger