Due to its ubiquitous availability, Short Message Service (SMS), first introduced in the 1980s, became not only the most popular way of communication, but also stimulated the development of SMS-based value added services. This application-to-person traffic is delivered to end users through SMS aggregators who provide the link between service providers and mobile carriers. In order to perform optimal traffic routing, the aggregators need to estimate the quality of each potential international route to the specified destination. The evaluation criteria include end-to-end delivery time, as well as correct verification of delivered data. This thesis suggests a method of quality of service (QoS) assessment for international SMS service which combines two types of tests, end-to-end delay measurements and various verification tests. A prototype of the testing system for international SMS service was developed to generate SMS traffic, collect and analyze results, and evaluate the experienced QoS of the SMS route used in accordance with the proposed approach. As a part of end-to- end delay measurement tests, SMS traffic was sent to Singtel network in Singapore along two routes. The verification tests were executed via different routes to two mobile networks: Singtel and Tele2 (Sweden). The results of the performed measurements determined the route with the highest QoS, i.e. the one with bigger bottleneck bandwidth and lower data loss rate. The prototype of the SMS testing system can be used by SMS aggregators to verify delivery of a SMS message, check the integrity of the message, figure out interconnection type of the route supplier with the destination carrier and to identify the presence of load balancers in the path. The prototype also makes it possible to compare end-to-end delay times of several routes and compute bottleneck values for each of the tested routes.

La certification des réseaux avioniques requiert une maîtrise des délais de transmission des données. Cepednant, le multiplexage et le partage des ressource de communications dans des réseaux tels que l'AFDX (Avionics Full Duplex Switched Ethernet) rendent difficile le calcul d'un délai de bout en bout pire cas pour chaque flux. Des outils comme le calcul réseau fournissent une borne supérieure (pessimiste) de ce délai pire cas. Les besoins de communication des avions civils modernes ne cessent d'augmenter et un nombre croissant de flux aux contraintes et aux caractéristiques différentes doivent partager les ressources existantes. Le réseau AFDX actuel ne permet pas de différentier plusieurs classes de trafic : les messages sont traités dans les files des commutateurs selon leur ordre d'arrivée (politique de service FIFO). L'objet de cette thèse est de montrer qu'il est possible de calculer des bornes pire cas des délais de bout en bout avec des politiques de service plus évoluées, à base de priorités statiques (Priority Queueing) ou à répartition équitable de service (Fair Queueing). Nous montrons comment l'approche par trajectoires, issue de la théorie de l'ordonnancement dans des systèmes asynchrones distribués peut s'appliquer au domaine de l'AFDX actuel et futur (intégration de politiques de service plus évoluées permettant la différentiation de flux). Nous comparons les performances de cette approche avec les outils de référence lorsque cela est possible et étudions le pessimisme des bornes ainsi obtenues. / The certification process for avionics network requires guaranties on data transmission delays. However, calculating the worst case delay can be complex in the case of industrial AFDX (Avionics Full Duplex Switched Ethernet) networks. Tools such as Network Calculus provide a pessimistic upper bound of this worst case delay. Communication needs of modern commercial aircraft are expanding and a growing number of flows with various constraints and characteristics must share already existing resources. Currently deployed AFDX networks do not differentiate multiple classes of traffic: messages are processed in their arrival order in the output ports of the switches (FIFO servicing policy). The purpose of this thesis is to show that it is possible to provide upper bounds of end to end transmission delays in networks that implement more advanced servicing policies, based on static priorities (Priority Queuing) or on fairness (Fair Queuing). We show how the trajectory approach, based on scheduling theory in asynchronous distributed systems can be applied to current and future AFDX networks (supporting advanced servicing policies with flow differentiation capabilities). We compare the performance of this approach with the reference tools whenever it is possible and we study the pessimism of the computed upper bounds.

In many practical problems, both decision and estimation are involved. This dissertation intends to study the relationship between decision and estimation in these problems, so that more accurate inference methods can be developed. Hybrid estimation is an important formulation that deals with state estimation and model structure identification simultaneously. Multiple-model (MM) methods are the most widelyused tool for hybrid estimation. A novel approach to predict the Internet end-to-end delay using MM methods is proposed. Based on preliminary analysis of the collected end-to-end delay data, we propose an off-line model set design procedure using vector quantization (VQ) and short-term time series analysis so that MM methods can be applied to predict on-line measurement data. Experimental results show that the proposed MM predictor outperforms two widely used adaptive filters in terms of prediction accuracy and robustness. Although hybrid estimation can identify model structure, it mainly focuses on the estimation part. When decision and estimation are of (nearly) equal importance, a joint solution is preferred. By noticing the resemblance, a new Bayes risk is generalized from those of decision and estimation, respectively. Based on this generalized Bayes risk, a novel, integrated solution to decision and estimation is introduced. Our study tries to give a more systematic view on the joint decision and estimation (JDE) problem, which we believe the work in various fields, such as target tracking, communications, time series modeling, will benefit greatly from. We apply this integrated Bayes solution to joint target tracking and classification, a very important topic in target inference, with simplified measurement models. The results of this new approach are compared with two conventional strategies. At last, a surveillance testbed is being built for such purposes as algorithm development and performance evaluation. We try to use the testbed to bridge the gap between theory and practice. In the dissertation, an overview as well as the architecture of the testbed is given and one case study is presented. The testbed is capable to serve the tasks with decision and/or estimation aspects, and is helpful for the development of the JDE algorithms.

Dans cette thèse, nous proposons une nouvelle approche pour modéliser et estimer les délais de bout-en-bout dans les réseaux de capteurs sans-fil (WSN). Notre approche combine les approches analytiques et expérimentales pour inférer un modèle Markovien modélisant le comportement d'un protocole de contrôle d'accès au médium (MAC) exécuté sur les nœuds d'un réseau de capteurs. À partir de ce modèle Markovien, le délai de bout en bout est ensuite obtenu par une approche analytique basée sur une analyse dans le domaine fréquentiel pour calculer la probabilité de distribution de délais pour un taux d'arrivée spécifique. Afin d’obtenir une estimation du délai de bout en bout, indépendamment du trafic en entrée, la technique de régression non-linéaire est utilisée sur un ensemble d’échantillons limités. Cette approche nous a permis de contourner deux problèmes: 1) la difficulté d'obtenir un modèle Markovien du comportement d’un protocole MAC en tenant compte de son implémentation réelle, 2) l'estimation du délai de bout-en-bout d’un WSN multi-sauts. L'approche a été validée sur un testbed réel (IOT-LAB) et pour plusieurs protocoles (X-MAC, ContikiMAC, IEEE 802.15.4) ainsi que pour un protocole de routage (RPL). / In this thesis, we propose an approach that combines both measurements and analytical approaches for infering a Markov chain model from the MAC protocol execution traces in order to be able to estimate the end to end delay in multi-hop transmission scenarios. This approach allows capturing the main features of WSN. Hence, a suitable Markov chain for modelling the WSN is infered. By means of an approach based on frequency domain analysis, end to end delay distribution for multi-hop scenarios is found. This is an important contribution of our approach with regard to existing analytical approaches where the extension of these models for considering multi-hop scenarios is not possible due to the fact that the arrival distribution to intermediate nodes is not known. Since local delay distribution for each node is obtained by analysing the MAC protocol execution traces for a given traffic scenario, the obtained model (and therefore, the whole end to end delay distribution) is traffic-dependant. In order to overcome this problem, we have proposed an approach based on non-linear regression techniques for generalising our approach in terms of the traffic rate. Results were validated for different MAC protocols (X-MAC, ContikiMAC, IEEE 802.15.4) as well as a well-known routing protocol (RPL) over real test-beds (IOT-LAB).

Depuis l'avènement du réseau Internet, le volume de données échangées sur les réseaux a crû de manière exponentielle. Le matériel présent sur les réseaux est devenu très hétérogène, dû entre autres à la multiplication des "middleboxes" (parefeux, routeurs NAT, serveurs VPN, proxy, etc.). Les algorithmes exécutés sur les équipements réseaux (routage, “spanning tree”, etc.) sont souvent complexes, parfois fermés et propriétaires et les interfaces de supervision peuvent être très différentes d'un constructeur/équipement à un autre. Ces différents facteurs rendent la compréhension et le fonctionnement du réseau complexe. Cela a motivé la définition d'un nouveau paradigme réseaux afin de simplifier la conception et la gestion des réseaux : le SDN (“Software-defined Networking”). Il introduit la notion de contrôleur, qui est un équipement qui a pour rôle de contrôler les équipements du plan de données. Le concept SDN sépare donc le plan de données chargés de l'acheminement des paquets, qui est opéré par des équipements nommés virtual switches dans la terminologie SDN, et le plan contrôle, en charge de toutes les décisions, et qui est donc effectué par le contrôleur SDN. Pour permettre au contrôleur de prendre ses décisions, il doit disposer d'une vue globale du réseau. En plus de la topologie et de la capacité des liens, des critères de performances comme le délai, le taux de pertes, la bande passante disponible, peuvent être pris en compte. Cette connaissance peut permettre par exemple un routage multi-classes, ou/et garantir des niveaux de qualité de service. Les contributions de cette thèse portent sur la proposition d'algorithmes permettant à une entité centralisée, et en particulier à un contrôleur dans un cadre SDN, d'obtenir des estimations fiables du délai de bout-en-bout pour les flux traversant le réseau. Les méthodes proposées sont passives, c'est-à-dire qu'elles ne génèrent aucun trafic supplémentaire. Nous nous intéressons tout particulièrement à la moyenne et l'écart type du délai. Il apparaît que le premier moment peut être obtenu assez facilement. Au contraire, la corrélation qui apparaît dans les temps d'attentes des noeuds du réseau rend l'estimation de l'écart type beaucoup plus complexe. Nous montrons que les méthodes développées sont capables de capturer les corrélations des délais dans les différents noeuds et d'offrir des estimations précises de l'écart type. Ces résultats sont validés par simulations où nous considérons un large éventail de scénarios permettant de valider nos algorithmes dans différents contextes d'utilisation / Since the early beginning of Internet, the amount of data exchanged over the networks has exponentially grown. The devices deployed on the networks are very heterogeneous, because of the growing presence of middleboxes (e.g., firewalls, NAT routers, VPN servers, proxy). The algorithms run on the networking devices (e.g., routing, spanning tree) are often complex, closed, and proprietary while the interfaces to access these devices typically vary from one manufacturer to the other. All these factors tend to hinder the understanding and the management of networks. Therefore a new paradigm has been introduced to ease the design and the management of networks, namely, the SDN (Software-defined Networking). In particular, SDN defines a new entity, the controller that is in charge of controlling the devices belonging to the data plane. Thus, in a SDN-network, the data plane, which is handled by networking devices called virtual switches, and the control plane, which takes the decisions and executed by the controller, are separated. In order to let the controller take its decisions, it must have a global view on the network. This includes the topology of the network and its links capacity, along with other possible performance metrics such delays, loss rates, and available bandwidths. This knowledge can enable a multi-class routing, or help guarantee levels of Quality of Service. The contributions of this thesis are new algorithms that allow a centralized entity, such as the controller in an SDN network, to accurately estimate the end-to-end delay for a given flow in its network. The proposed methods are passive in the sense that they do not require any additional traffic to be run. More precisely, we study the expectation and the standard deviation of the delay. We show how the first moment can be easily computed. On the other hand, estimating the standard deviation is much more complex because of the correlations existing between the different waiting times. We show that the proposed methods are able to capture these correlations between delays and thus providing accurate estimations of the standard deviation of the end-to-end delay. Simulations that cover a large range of possible scenariosvalidate these results

L’importance des systèmes temps-réels embarqués dans les sociétés industrialisées modernes en font un terrain d’application privilégié pour les méthodes formelles. La prépondérance des contraintes temporelles dans les spécifications de ces systèmes motive la mise au point de solutions spécifiques. Cette thèse s’intéresse à une classe de systèmes temps-réels incluant ceux développés avec la chaîne d’outils OASIS, développée au CEA LIST. Nos travaux portent sur la notion de délai de bout-en-bout, que nous proposons de modéliser comme une contrainte temporelle concernant l’influence du flot d’informations des entrées sur celui des sorties. Afin de répondre à la complexité croissante des systèmes temps-réels, nous étudions l’applicabilité de cette notion nouvelle au développement incrémental par raffinement et par composition. Le raffinement est abordé sous l’angle de la conservation de propriétés garantes de la correction du système au cours du processus de développement. Nous délimitons les conditions nécessaires et suffisantes à la conservation du délai de bout-en-bout lors d’un tel processus. De même, nous donnons des conditions suffisantes pour permettre le calcul du délai de bout-en-bout de manière compositionnelle. Combinés, ces résultats permettent d’établir un formalisme permettant la preuve du délai de bout-en-bout lors d’une démarche de développement incrémentale. / Real-time embedded systems are at the core of modern industrialized societies. They are a privileged target for the application of formal methods. The importance of real-time constraints in the specification of these systems requires the design of ad-hoc solutions. This work considers a class of real-time systems including those developed using OASIS, a tool-chain targeting hard real-time embedded systems developed at CEA LIST. We study the notion of end-to-end delay, which we propose to model as a constraint bearing directly on the influence of the input information flow over the output information flow . In order to cope with the growing complexity of real-time embedded systems, we study the possibility to apply this new notion of delay to the incremental development of such systems, by using both stepwise refinement and composition operators. We define the necessary and sufficient conditions to the preservation of the end-to-end delay by stepwise refinement. Similarly, we give sufficient conditions to compute the end-to-end delay in a compositional fashion. Together, these results permit to establish a formalism allowing to prove end-to-end delay properties in stepwise development methodologies.

Timeliness in delivering packets for delay-sensitive applications is an important QoS (Quality of Service) measure in many systems, notably those that need to provide real-time performance. In such systems, if delay-sensitive traffic is delivered to the destination beyond the deadline, then the packets will be rendered useless and dropped after received at the destination. Bandwidth that is already scarce and shared between network nodes is wasted in relaying these expired packets. This thesis proposes that a deterministic per-hop delay can be achieved by using a dynamic queue threshold concept to bound delay of each node. A deterministic per-hop delay is a key component in guaranteeing a deterministic end-to-end delay. The research aims to develop a generic approach that can constrain network delay of delay-sensitive traffic in a dynamic network. Two adaptive queue management schemes, namely, DTH (Dynamic THreshold) and ADTH (Adaptive DTH) are proposed to realize the claim. Both DTH and ADTH use the dynamic threshold concept to constrain queuing delay so that bounded average queuing delay can be achieved for the former and bounded maximum nodal delay can be achieved for the latter. DTH is an analytical approach, which uses queuing theory with superposition of N MMBP-2 (Markov Modulated Bernoulli Process) arrival processes to obtain a mapping relationship between average queuing delay and an appropriate queuing threshold, for queue management. While ADTH is an measurement-based algorithmic approach that can respond to the time-varying link quality and network dynamics in wireless ad hoc networks to constrain network delay. It manages a queue based on system performance measurements and feedback of error measured against a target delay requirement. Numerical analysis and Matlab simulation have been carried out for DTH for the purposes of validation and performance analysis. While ADTH has been evaluated in NS-2 simulation and implemented in a multi-hop wireless ad hoc network testbed for performance analysis. Results show that DTH and ADTH can constrain network delay based on the specified delay requirements, with higher packet loss as a trade-off.

In critical real-time systems, any faulty behavior may endanger lives. Hence, system verification and validation is essential before their deployment. In fact, safety authorities ask to ensure deterministic guarantees. In this thesis, we are interested in offering temporal guarantees; in particular we need to prove that the end-to-end response time of every flow present in the network is bounded. This subject has been addressed for many years and several approaches have been developed. After a brief comparison between the existing approaches, the Trajectory Approach sounded like a good candidate due to the tightness of its offered bound. This method uses results established by the scheduling theory to derive an upper bound. The reasons leading to a pessimistic upper bound are investigated. Moreover, since the method must be applied on large networks, it is important to be able to give results in an acceptable time frame. Hence, a study of the method's scalability was carried out. Analysis shows that the complexity of the computation is due to a recursive and iterative processes. As the number of flows and switches increase, the total runtime required to compute the upper bound of every flow present in the network understudy grows rapidly. While based on the concept of the Trajectory Approach, we propose to compute an upper bound in a reduced time frame and without significant loss in its precision. It is called the Scalable Trajectory Approach. After applying it to a network, simulation results show that the total runtime was reduced from several days to a dozen seconds.

Peer to peer networks are becoming increasingly popular among Internet users as the downloading peers share the storage and upload bandwidth load of the system. This makes it possible for a large number of users to share a data file available at a server without the server upload bandwidth becoming a bottleneck. The P2P technology is being widely used not only for file sharing but also for video on demand, live streaming and IPTV. The delay deadlines are more stringent in live streaming and IPTV than those in file sharing as the traffic is real time. The performance perceived by a user depends upon whether the video stream is being downloaded at the streaming rate. Coolstreaming is the first large scale P2P IPTV system. We model the multi-channel Coolstreaming system via an open queueing network. The peer dynamics at a channel is modelled by a closed queueing network working at a faster rate. We compute the expected number of substreams in the overlay of New Coolstreaming which are not being received at the proper rate. The computation of the Markov chain with a very large state space is handled using the two time scale decomposition. Further we characterize the end to end delay encountered by a video stream originating from the server and received at a user of New Coolstreaming. Three factors contribute towards the delay. The first factor is the mean path length in terms of overlay hops of the partnership graph. The second factor is the mean number of routers between any two overlay peers in the network layer and the third factor is the queueing delay at a router in the Internet. The mean shortest path length in terms of overlay peers in the New Coolstreaming graph is shown to be O(logn)where nis the number of peers in the overlay. This is done by modelling the overlay by a random graph. The mean shortest path in terms of routers in the Internet’s router level topology is seen to be at most O(logNI)where NIis the number of routers in the Internet. We also discuss a method by which we can get the mean delay at a router in the Internet. Thus, the mean end to end delay in New Coolstreaming is shown to be upper bounded by O(lognlogNIE[W])where E[W]is the mean delay at a router in the Internet.

Software systems in the automotive domain are generally safety critical and subject to strict timing requirements. Systems of this character are often constructed utilizing periodically executed tasks, that have a hard deadline. In addition, these systems may have additional deadlines that can be specified on cause-effect chains, or simply task chains. They are defined by existing tasks in the system, hence the chains are not stand alone additions to the system. Each chain provide an end-to-end timing constraint targeting the propagation of data through the chain of tasks. These constraints specify the additional timing requirements that need to be fulfilled, when searching for a valid schedule. In this thesis, an offline non-preemptive scheduling method is presented, designed for single core systems. The scheduling problem is defined and formulated utilizing Constraint Programming. In addition, to ensure that end-to-end timing requirements are met, job-level dependencies are considered during the schedule generation. Utilizing this approach can guarantee that individual task periods along with end-to-end timing requirements are always met, if a schedule exists. The results show a good increase in schedulability ratio when utilizing job-level dependencies compared to the case where job-level dependencies are not specified. When the system utilization increases this improvement is even greater. Depending on the system size and complexity the improvement can vary, but in many cases it is more than double. The scheduling generation is also performed within a reasonable time frame. This would be a good benefit during the development process of a system, since it allows fast verification when changes are made to the system. Further, the thesis provide an overview of the entire process, starting from a system model and ending at a fully functional schedule executing on a hardware platform.