Previous research has developed a number of efficient protocols for streaming popular multimedia files on-demand to potentially large numbers of concurrent clients. These protocols can achieve server bandwidth usage that grows much slower than linearly with the file request rate, and with the inverse of client start-up delay. This hesis makes the following three main contributions to the design and performance evaluation of such protocols. The first contribution is an investigation of the network bandwidth requirements for scalable on-demand streaming. The results suggest that the minimum required network bandwidth for scalable on-demand streaming typically scales as K/ln(K) as the number of client sites K increases for fixed request rate per client site, and as ln(N/(ND+1)) as the total file request rate N increases or client start-up delay D decreases, for a fixed number of sites. Multicast delivery trees configured to minimize network bandwidth usage rather than latency are found to only modestly reduce the minimum required network bandwidth. Furthermore, it is possible to achieve close to the minimum possible network and server bandwidth usage simultaneously with practical scalable delivery protocols. Second, the thesis addresses the problem of scalable on-demand streaming of a more complex type of media than is typically considered, namely variable bit rate (VBR) media. A lower bound on the minimum required server bandwidth for scalable on-demand streaming of VBR media is derived. The lower bound analysis motivates the design of a new immediate service protocol termed VBR bandwidth skimming (VBRBS) that uses constant bit rate streaming, when sufficient client storage space is available, yet fruitfully exploits the knowledge of a VBR profile. Finally, the thesis proposes non-linear media containing parallel sequences of data frames, among which clients can dynamically select at designated branch points, and investigates the design and performance issues in scalable on-demand streaming of such media. Lower bounds on the minimum required server bandwidth for various non-linear media scalable on-demand streaming approaches are derived, practical non-linear media scalable delivery protocols are developed, and, as a proof-of-concept, a simple scalable delivery protocol is implemented in a non-linear media streaming prototype system.

Previous research has developed a number of efficient protocols for streaming popular multimedia files on-demand to potentially large numbers of concurrent clients. These protocols can achieve server bandwidth usage that grows much slower than linearly with the file request rate, and with the inverse of client start-up delay. This hesis makes the following three main contributions to the design and performance evaluation of such protocols. The first contribution is an investigation of the network bandwidth requirements for scalable on-demand streaming. The results suggest that the minimum required network bandwidth for scalable on-demand streaming typically scales as K/ln(K) as the number of client sites K increases for fixed request rate per client site, and as ln(N/(ND+1)) as the total file request rate N increases or client start-up delay D decreases, for a fixed number of sites. Multicast delivery trees configured to minimize network bandwidth usage rather than latency are found to only modestly reduce the minimum required network bandwidth. Furthermore, it is possible to achieve close to the minimum possible network and server bandwidth usage simultaneously with practical scalable delivery protocols. Second, the thesis addresses the problem of scalable on-demand streaming of a more complex type of media than is typically considered, namely variable bit rate (VBR) media. A lower bound on the minimum required server bandwidth for scalable on-demand streaming of VBR media is derived. The lower bound analysis motivates the design of a new immediate service protocol termed VBR bandwidth skimming (VBRBS) that uses constant bit rate streaming, when sufficient client storage space is available, yet fruitfully exploits the knowledge of a VBR profile. Finally, the thesis proposes non-linear media containing parallel sequences of data frames, among which clients can dynamically select at designated branch points, and investigates the design and performance issues in scalable on-demand streaming of such media. Lower bounds on the minimum required server bandwidth for various non-linear media scalable on-demand streaming approaches are derived, practical non-linear media scalable delivery protocols are developed, and, as a proof-of-concept, a simple scalable delivery protocol is implemented in a non-linear media streaming prototype system.

Ce travail s'inscrit dans le cadre de la vérification formelle de programmes. La vérification de modèle permet de s'assurer qu'une propriété est vérifiée par le modèle du système. Cette thèse étudie la vérification paramétrée de réseaux composés d'un nombre non borné de processus identiques où le nombre de processus est considéré comme un paramètre. Concernant les réseaux de protocoles probabilistes temporisés nous montrons que les problèmes de l'accessibilité et de synchronisation sont indécidables pour des topologies de communication en cliques. Cependant, en considérant des pertes et créations probabiliste de processus ces problèmes deviennent décidables. Pour ce qui est des réseaux dans lequel les messages n'atteignent qu'une sous partie des composants choisie de manière non-déterministe, nous prouvons que le problème de l'accessibilité paramétrée est décidable grâce à une réduction à un nouveau modèle de jeux à deux joueurs distribué pour lequel nous montrons que l'on peut décider de l'existence d'une stratégie gagnante en coNP. Finalement, nous considérons des stratégies locales qui permettent d'assurer que les processus effectuent leurs choix non-déterministes uniquement par rapport a leur connaissance locale du système. Sous cette hypothèse de stratégies locales, nous prouvons que les problèmes de l'accessibilité et de synchronisation paramétrées sont NP-complet. / This thesis deals with formal verification of distributed systems. Model checking is a technique for verifying that the model of a system under study fulfills a given property. This PhD investigates the parameterized verification of networks composed of many identical processes for which the number of processes is the parameter. Considering networks of probabilistic timed protocols, we show that the parameterized reachability and synchronization problems are undecidable when the communication topology is a clique. However, assuming probabilistic creation and deletion of processes, the problems become decidable. Regarding selective networks, where the messages only reach a subset of the components, we show decidability of the parameterized reachability problem thanks to reduction to a new model of distributed two-player games for which we prove decidability in coNP of the game problem. Finally, we consider local strategies that enforce all processes to resolve the non-determinism only according to their own local knowledge. Under this assumption of local strategy, we were able to show that the parameterized reachability and synchronization problems are NP-complete.

An ad hoc wireless network is an infrastructure-less, autonomous system of nodes connected through wireless links. In many current applications of ad hoc wireless networks, individual wireless nodes are autonomous, rational, and intelligent and are often referred to as selfish nodes, following game theoretic terminology. In an ad hoc wireless network, a typical node may be an intermediate node of a route from a source node to a destination node and therefore is often required to forward packets so as to enable communication to be established. Selfish nodes may not always forward the packets since the forwarding activity consumes the node’s own resources. Such behavior by individual nodes may lead to suboptimal situations where nodes, through their actions, lead to a state that is undesirable from an overall network viewpoint. To counter this, there is a need to stimulate cooperation through methods such as providing appropriate incentives. In this thesis, our interest is in designing rigorous incentive based methods for stimulating cooperation among wireless nodes, in the specific context of broadcast. In particular, we address the Incentive Compatible Broadcast problem: how do we design broadcast protocols that induce truth revelation by the individual wireless nodes? We do this using a game theory and mechanism design framework. Incentive compatibility of broadcast protocols could manifest in two forms: (1) Dominant Strategy Incentive Compatibility (DSIC) (also called strategy-proofness) and (2) Bayesian incentive compatibility (BIC). A DSIC broadcast protocol is one which makes it a best response for every wireless node to reveal its true type, regardless of what the other nodes reveal. A BIC broadcast protocol is one which makes truth revelation a best response for a node, given that the other nodes are truthful. The DSIC property is stronger and more desirable but more difficult to achieve. On the other hand, the BIC property is much weaker and easier to achieve. In this thesis, we first design a DSIC broadcast protocol for ad hoc networks using the well known VCG (Vickrey-Clarke-Groves) mechanisms and investigate its properties and performance. Next, we design a BIC broadcast protocol, investigate its properties, and compare its performance with that of the DSIC broadcast protocol. Both the protocols developed in this thesis provide an elegant solution to the incentive compatible broadcast problem in ad hoc networks with selfish nodes and help stimulate cooperation among the selfish wireless nodes.

De nos jours, la voiture est devenue le mode de transport le plus utilisé, mais malheureusement, il est accompagné d’un certain nombre de problèmes (accidents, pollution, embouteillages, etc.), qui vont aller en s’aggravant avec l’augmentation prévue du nombre de voitures particulières, malgré les efforts très importants mis en œuvre pour tenter de les réduire ; le nombre de morts sur les routes demeure très important. Les réseaux sans fil de véhicules, appelés VANET, qui consistent de plusieurs véhicules mobiles sans infrastructure préexistante pour communiquer, font actuellement l’objet d'une attention accrue de la part des constructeurs et des chercheurs, afin d’améliorer la sécurité sur les routes ou encore les aides proposées aux conducteurs. Par exemple, ils peuvent avertir d’autres automobilistes que les routes sont glissantes ou qu’un accident vient de se produire. Dans VANET, les protocoles de diffusion (broadcast) jouent un rôle très important par rapport aux messages unicast, car ils sont conçus pour transmettre des messages de sécurité importants à tous les nœuds. Ces protocoles de diffusion ne sont pas fiables et ils souffrent de plusieurs problèmes, à savoir : (1) Tempête de diffusion (broadcast storm) ; (2) Nœud caché (hidden node) ; (3) Échec de la transmission. Ces problèmes doivent être résolus afin de fournir une diffusion fiable et rapide. L’objectif de notre recherche est de résoudre certains de ces problèmes, tout en assurant le meilleur compromis entre fiabilité, délai garanti, et débit garanti (Qualité de Service : QdS). Le travail de recherche de ce mémoire a porté sur le développement d’une nouvelle technique qui peut être utilisée pour gérer le droit d’accès aux médias (protocole de gestion des émissions), la gestion de grappe (cluster) et la communication. Ce protocole intègre l'approche de gestion centralisée des grappes stables et la transmission des données. Dans cette technique, le temps est divisé en cycles, chaque cycle est partagé entre les canaux de service et de contrôle, et divisé en deux parties. La première partie s’appuie sur TDMA (Time Division Multiple Access). La deuxième partie s’appuie sur CSMA/CA (Carrier Sense Multiple Access / Collision Avoidance) pour gérer l’accès au medium. En outre, notre protocole ajuste d’une manière adaptative le temps consommé dans la diffusion des messages de sécurité, ce qui permettra une amélioration de la capacité des canaux. Il est implanté dans la couche MAC (Medium Access Control), centralisé dans les têtes de grappes (CH, cluster-head) qui s’adaptent continuellement à la dynamique des véhicules. Ainsi, l’utilisation de ce protocole centralisé nous assure une consommation efficace d’intervalles de temps pour le nombre exact de véhicules actifs, y compris les nœuds/véhicules cachés; notre protocole assure également un délai limité pour les applications de sécurité, afin d’accéder au canal de communication, et il permet aussi de réduire le surplus (overhead) à l’aide d’une propagation dirigée de diffusion. / Nowadays, the car has become the most popular mode of transport, but unfortunately its use is accompanied by a number of problems (accidents, pollution, congestion, etc.). These problems will get worse with the increase in the number of passenger cars, despite very significant efforts made to reduce the number of road deaths, which is still very high. Wireless networks for vehicles called VANET (Vehicle Ad Hoc Networks), were developed when it became possible to connect several mobile vehicles without relying on pre existing communication infrastructures. These networks have currently become the subject of increased attention from manufacturers and researchers, due to their potential for improving road safety and/or offering assistance to drivers. They can, for example, alert other drivers that roads are slippery or that an accident has just occurred. In VANETs, broadcast protocols play a very important role compared to unicast protocols, since they are designed to communicate important safety messages to all nodes. Existing broadcast protocols are not reliable and suffer from several problems: (1) broadcast storms, (2) hidden nodes, and (3) transmission failures. These problems must solved if VANETs are to become reliable and able to disseminate messages rapidly. The aim of our research is to solve some of these problems while ensuring the best compromise among reliability, guaranteed transmission times and bandwidth (Quality of Service: QoS). The research in this thesis focuses on developing a new technique for managing medium access. This protocol incorporates the centralized management approach involving stable clusters. In this technique, time is divided into cycles; with each cycle being shared among the control and service channels, and is divided into two segments. The first is based on TDMA (Time Division Multiple Access) while the second is based on CSMA/CA (Carrier Sense Multiple Access/Collision Avoidance) to manage access to the medium. Furthermore, our protocol adaptively adjusts the time consumed in broadcasting safety messages, thereby improving channel capacity. It is implemented in the MAC (Medium Access Control), and centralized in stable cluster heads that are able to adapt to the dynamics of vehicles. This protocol provides a centralized and efficient use of time intervals for an exact number of active vehicles, including hidden nodes/vehicles. Our protocol also provides time intervals dedicated to security applications for providing access to communication channels, and also reduces overhead via directed diffusion of data. Keywords: Ad-hoc networks, VANET, Vehicle, Periodic Safety Messages, broadcast protocols, contention-free.

No description available.